Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans

Andre Luiz Freire de Carvalho Kato; Auren Liu; Chayapatr Archiwaranguprok; Constanze Albrecht; Emillie de Keulenaar; Jackson Trager; Jennifer Pfister; Juliana Schroeder; Krystal Kirkland; Nathanael J. Fast

REVIEW

Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans

Human-Computer Interaction · Submitted 14 Oct 2026 · 151 references

Pith's verdict: A massive, well-executed benchmarking infrastructure for AI human impact, but the judge-model conflict is a load-bearing flaw that needs addressing before the numerical results can be fully trusted.

T0 · AI review

How this review was made: An AI referee (deepseek/deepseek-v4-flash) read the paper on October 16, 2026 and answered a fixed list of questions about it. T0 is the starting level every paper gets. The levels describe how deeply Pith has checked the paper, never the authors.

T0 AI review An AI read the paper and answered a fixed list of questions.
T1 Audit Automatic checks that the references and files the paper relies on exist and match.
T2 Compute The paper's code or calculations re-run with the same results.
T3 Formal statement The claims written out in a formal math language a computer can read.
T4 Proof Those claims proved, and the proof checked by computer.

How Pith checks papers →

· 3 major objections · 5 minor · Ask about this paper · The compiled record · Read the full review →

Challenge this review Comment (0) Read on arXiv ↗

Cite ▾

arXiv ID2610.20690 (version 1)

Pith IDQNKOGC6N, Pith's permanent record of this paper and its review

CitationAndre Luiz Freire de Carvalho Kato, Auren Liu, Chayapatr Archiwaranguprok, Constanze Albrecht, Emillie de Keulenaar, Jackson Trager, Jennifer Pfister, Juliana Schroeder, Krystal Kirkland, Nathanael J. Fast, Nikos Bellos, Pat Pataranutaporn, Pattie Maes, Peter Baxter, Rachel Poonsiriwong, Ravi Iyer, Sheer Karny, Stanley Huang, Valdemar Danry, Yuning Liu. "Open Benchmark of AI Impact on Humans (ImpactBench): an expert-guided, open platform for the holistic evaluation of AI impact on humans." arXiv:2610.20690 (2026). https://pith.science/paper/QNKOGC6N

Cite this reviewBibTeX for Pith's review of this paper

Claim paper

Why Pith says this

Highlight any text to comment on it.

Editor's note

read the letter

You should know this paper is a serious infrastructure effort: 809 metrics across 26 benchmarks, 10 models, 48k simulated conversations, with multi-turn design, expert validation, sensitivity analyses, and a public platform. That is real work. The central empirical claim—that models are better at promoting positive behaviors (68.7%) than avoiding negative ones (53.1%)—is likely robust as a qualitative pattern, because it holds across all ten models. The paper also gives credit where due: internal and external expert reviews, inter-judge reliability checks, ablation of anchors, and a candid limitations section that flags that these are behavioral proxies and that bootstrap intervals condition on saved judgments.

Now the soft spot that matters most. The judge model (GPT-5.6 Luna) is also a target model. The same checkpoint family scores itself and its sibling (Terra). This is not a minor oversight—it is a potential systematic bias that could inflate Luna's pass rates relative to models from other families. The stress-test note is correct: the paper provides no analysis of whether the judge favors its own family. The human validation on 18 metrics (297 ratings) did not test for family-specific bias. Because the paper’s central findings include model rankings and the claim that "negative metrics separate models far more than positive metrics," the specific percentages and between-model differences become uninterpretable without a judge-impartiality check. The positive–negative gap might survive a bias correction, but the paper does not show that.

Other concerns are milder. Inter-reviewer agreement was low on some metrics (Krippendorff's alpha 0.121 for scaffolded learning), which the paper acknowledges. The bootstrap intervals do not capture variance from the simulator or the judge, also acknowledged. So the main unresolved issue is the judge-model conflict.

This paper deserves a serious referee: the infrastructure is valuable and the methodology is mostly sound. But a referee should push for an analysis of potential judge bias—for example, re-evaluating a subset with a judge from a different family, or at minimum a systematic discussion of why the conflict might not matter. Who benefits? Anyone building or using human-impact evaluations, including policymakers, developers, and researchers. I would cite the paper for its infrastructure design, not for the specific numerical results until the conflict is resolved.

Referee report

Significance

If the results hold, ImpactBench provides a much-needed infrastructure for holistic, transparent evaluation of AI's human impact, addressing structural gaps in current benchmarking practice: who defines constructs, how rigorously they are measured, whether evaluation captures multi-turn behavior, and how results are communicated beyond the technical community. The scale (48,540 conversations, 809 metrics, 10 models), the inclusion of multi-turn simulation and matched teen/adult scenarios, the human validation effort, and the sensitivity analyses are notable strengths. The framework is designed to evolve through open submissions, which is important for a contested and rapidly developing area. The central empirical finding that models foster positive outcomes more reliably than they mitigate negative ones, and that negative metrics separate models far more than positive ones, is consequential if substantiated.

Major objections (3)

[§3.5.2, §3.4.6] Judge-model conflict: GPT-5.6 Luna serves as both the judge (§3.5.2) and one of the ten target models (§3.4.6). The paper does not evaluate whether the judge systematically favors its own family or checkpoint. This could inflate Luna's pass rates and affect the central claims about model rankings and the magnitude of the positive–negative gap (e.g., §4.4 reports Luna's positive rate 74.9%, negative 66.3%). The human validation (§4.9) covers multiple target models but does not report agreement rates broken down by model family, and the bootstrap intervals (§4.10) condition on saved judgments and thus do not capture judge-induced systematic error. The authors should either (a) provide evidence that the judge's impartiality across model families is adequate (e.g., human-model agreement rates per family, or a re-evaluation with a different judge on a subset), or (b) explicitly discuss this limitation and its potential impact on the conclusions about model rankings and the positive–negative asymmetry.
[§4.9] Human validation coverage: Expert review covered only 18 of 809 metrics (2.2%) and 297 of 48,540 conversations (0.6%), with only four metrics receiving overlapping reviews. The paper acknowledges this limitation (§5.6), but the central claim that the automated judge is reliable (83.3% majority agreement, Cohen's Kappa = 0.667) is based on this small, non-random sample. Given the low inter-reviewer agreement on some metrics (Krippendorff's alpha as low as 0.121 for scaffolded learning), the generalizability of the validation to the full metric suite is uncertain. The authors should clarify the sampling strategy for human review (was it random or stratified?) and provide a more precise statement about which kinds of metrics the validation supports. In particular, the validation should be analyzed for whether agreement varies by model family or metric polarity.
[§4.4, Table 3] The claim that 'negative metrics separate models far more than positive metrics' is supported by the range of pass rates (15.9 points for positive vs. 38.6 points for negative). However, this difference could be partly driven by the choice of metrics: negative metrics may be intrinsically more diverse or harder, and the paper does not control for metric difficulty or distribution. The authors should discuss whether the metric set is balanced in difficulty between polarities, or at least acknowledge that the separation may reflect metric selection rather than a fundamental property of model behavior. Additionally, the ranking correlation (Spearman ρ=0.94) suggests that ordering is similar across polarities, which somewhat tempers the claim about separation.

Minor objections (5)

[Figure 5] The color scale (red to green) is not explicitly defined; adding a legend would improve interpretability for readers, especially those who may view the figure in grayscale.
[§3.2.2] The list of 26 benchmarks is cited only by name and references. While Table 2 provides a useful summary, consider adding a brief one-sentence description of each benchmark's focus in the main text or appendix to help readers unfamiliar with each source.
[Appendix C, Table 4] The paper reports Krippendorff's alpha for four HumaneBench metrics based on 9 items per metric. Given the small sample size, the paper should explicitly caution that these coefficients have wide uncertainty and should not be over-interpreted. Also note that the table's caption states 'These evaluations used the earlier model set; they are not agreement estimates for the entire main production evaluation,' which is helpful but could be more prominent.
[Page 46, Table 1] The formatting of the table in the preprint shows slight misalignment in the column for Model names; ensure proper alignment in the camera-ready version for readability.
[§3.5.2, §3.3.2] The paper uses 'gpt-5.6-sol' for scenario generation and 'gpt-5.6-luna' as the judge. Clarify whether these are the same underlying model or different versions/checkpoints, and if possible report the exact version or checkpoint identifiers to support reproducibility.

How Pith checks papers → · Full audit of this paper →

The paper

The paper claims

ImpactBench evaluates 10 frontier LLMs across 809 metrics of human flourishing and finds a persistent gap: models promote good behaviors in 68.7% of cases but avoid harm in only 53.1%.

The reading

ImpactBench is an open, expert-guided platform that evaluates how generative AI models affect human well-being across psychological, physical, and social dimensions. The paper establishes a framework for translating domain expertise and existing benchmarks into 809 observable behavioral metrics, tested through 48,540 simulated multi-turn conversations with ten frontier LLMs. Its central claim is that current models are substantially better at promoting desirable behaviors (68.7% pass rate) than at avoiding harmful ones (53.1% pass rate), with particularly large gaps in learning support, autonomy preservation, and emotional boundaries. The authors argue that human-impact evaluation requires separating positive and negative behavioral criteria, multi-turn interaction testing, and open infrastructure that allows diverse stakeholders to contribute and inspect metrics.

What carries the argument

The ImpactBench evaluation pipeline converts expert-identified constructs and existing benchmark criteria into standardized behavioral metrics, each specifying a target behavior and a polarity (positive: desirable behavior present, or negative: harmful behavior absent). For each metric, a generator creates three base scenarios with demographic variants (teen and adult), which drive a user simulator (Claude Sonnet 5) through guided multi-turn conversations with the target model. A judge model (GPT-5.6 Luna) then determines whether the target behavior occurred in the conversation. The pipeline incorporates scenario quality checks, simulator realism ratings, judge repeatability tests, and human expert validation to assess measurement reliability.

Extended reading notes

Core claim

Across 48,540 simulated conversations covering 809 metrics from 26 benchmarks, ten frontier LLMs passed positive metrics (desirable behavior present) 68.7% of the time but passed negative metrics (harmful behavior absent) only 53.1% of the time. This asymmetry was largest in the psychological domain (24.2 percentage points) and most pronounced in subareas such as Learning & Skill Development, where models promoted learning behaviors in 45.9% of conversations but avoided undermining learning in only 16.3%. Models also differed far more on harm avoidance (range 29.8% to 68.4%) than on benefit promotion (59.2% to 75.1%). In teen contexts, models less often supported independent reasoning and more often displayed dependence-building behaviors than in adult contexts. External expert review found 83.3% agreement with the automated LLM judge (Cohen's kappa = 0.667), though inter-reviewer agreement on some metrics was low (Krippendorff's alpha as low as 0.121).

Load-bearing premise

The automated LLM judge (GPT-5.6 Luna) reliably identifies the presence or absence of target behaviors in conversations, and the human validation on 18 metrics (297 ratings) generalizes to the full suite of 809 metrics, despite low inter-reviewer agreement on some dimensions.

Editorial extensions

If this is right

Evaluations of AI systems should separately measure promotion of beneficial behaviors and avoidance of harmful ones, because these are distinct dimensions of model performance that can diverge.
Model rankings in human-impact benchmarks depend heavily on which metrics are included and how they are weighted; aggregate scores alone are misleading and should be reported alongside benchmark-level and metric-level results.
Developers should prioritize improving harm avoidance, especially in learning, autonomy, and emotional-boundary contexts, where current models exhibit the largest gaps.
Age-adapted evaluation reveals that models may pose unique risks to adolescent users, warranting targeted safety measures beyond those for adult users.
The open, modular design of ImpactBench enables diverse stakeholders to contribute metrics, making the evaluation ecosystem adaptable to evolving evidence and values.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the behavioral patterns observed in simulation translate to real-world interactions, users of current frontier AI systems are likely experiencing frequent subtle harms related to cognitive offloading, decision substitution, and emotional dependency, even when the models appear helpful.
The wide range in negative-metric performance across models suggests harm avoidance is a tractable engineering target; it may be possible to improve through targeted training without sacrificing helpfulness.
ImpactBench's modular open-submission design could serve as a template for regulatory frameworks requiring independent pre-deployment evaluation of AI impact on human flourishing, analogous to clinical trials for medical interventions.
The low inter-reviewer agreement on some metrics (e.g., scaffolded learning, alpha = 0.121) indicates a need for deeper conceptual work before those criteria are used for high-stakes decisions.

Original abstract

The present work is concerned with how to measure, at scale, the multi-dimensional impact of generative AI models on the individuals who use them. In service of this goal, we introduce the Open Benchmark of AI Impact on Humans, or ImpactBench. Our iterable and expert-guided platform evaluates whether model behaviors support or undermine an individual's wellbeing across psychological, physical, and social domains. ImpactBench translates constructs from domain experts and existing benchmarks into metrics that are grounded in actual model behavior. Success on these metrics occurs either when a model promotes expert-defined desirable behaviors or avoids identified harms. Models are evaluated against these metrics in simulated multi-turn conversations using matched teen and adult scenarios. The conversations are then scored by an LLM judge and verified through reliability audits and human expert review. The current suite delivers 809 metrics drawn from 26 benchmarks. We evaluated ten frontier LLMs against these metrics across 48,540 conversations. Our results indicate that models promoted desirable behaviors in 68.7% of evaluations, but avoided harms in only 53.1%. Majority expert ratings agreed with the automated judge in 83.3% of reviewed cases (Cohen's Kappa = 0.667). Results are released through a public platform that keep aggregate scores traceable to the underlying metrics and transcripts. ImpactBench is intended as evolving infrastructure for independent, transparent evaluation of generative AI's multi-dimensional impact on humans.

The compiled record

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-10-16: 1 request, 9,339 input tokens.

What the abstract itself states, as probabilities: claims first 0.06 · states limitation 0.30 · human participants 0.44 · controlled comparison 0.07 · randomized assignment 0.03 · states quantitative result 0.99 · compares with prior results 0.24 · mentions code or data availability 0.69.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; show every role.

The present work is concerned with how to measure, at scale, the multi-dimensional impact of generative AI models on the individuals who use them. In service of this goal, we introduce the Open Benchmark of AI Impact on Humans, or ImpactBench. Our iterable and expert-guided platform evaluates whether model behaviors support or undermine an individual's wellbeing across psychological, physical, and social domains. ImpactBench translates constructs from domain experts and existing benchmarks into metrics that are grounded in actual model behavior. Success on these metrics occurs either when a model promotes expert-defined desirable behaviors or avoids identified harms. Models are evaluated against these metrics in simulated multi-turn conversations using matched teen and adult scenarios. The conversations are then scored by an LLM judge and verified through reliability audits and human expert review. The current suite delivers 809 metrics drawn from 26 benchmarks. We evaluated ten frontier LLMs against these metrics across 48,540 conversations. Our results indicate that models promoted desirable behaviors in 68.7% of evaluations, but avoided harms in only 53.1%. Majority expert ratings agreed with the automated judge in 83.3% of reviewed cases (Cohen's Kappa = 0.667). Results are released through a public platform that keep aggregate scores traceable to the underlying metrics and transcripts. ImpactBench is intended as evolving infrastructure for independent, transparent evaluation of generative AI's multi-dimensional impact on humans.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#	sentence	role	the authors' own work	states the main conclusion	strength
s0	The present work is concerned with how to measure, at scale, the multi-dimensional impact of generative AI models on the individuals who use them.	background 0.93	0.16	0.03	asserted 0.77
s1	In service of this goal, we introduce the Open Benchmark of AI Impact on Humans, or ImpactBench.	contribution 1.00	0.87	0.02	asserted 0.98
s2	Our iterable and expert-guided platform evaluates whether model behaviors support or undermine an individual's wellbeing across psychological, physical, and social domains.	contribution 0.68	0.76	0.05	asserted 1.00
s3	ImpactBench translates constructs from domain experts and existing benchmarks into metrics that are grounded in actual model behavior.	method 0.90	0.84	0.03	asserted 1.00
s4	Success on these metrics occurs either when a model promotes expert-defined desirable behaviors or avoids identified harms.	method 0.89	0.22	0.05	asserted 1.00
s5	Models are evaluated against these metrics in simulated multi-turn conversations using matched teen and adult scenarios.	method 1.00	0.83	0.02	asserted 1.00
s6	The conversations are then scored by an LLM judge and verified through reliability audits and human expert review.	method 1.00	0.89	0.02	asserted 1.00
s7	The current suite delivers 809 metrics drawn from 26 benchmarks.	method 0.61	0.84	0.03	asserted 1.00
s8	We evaluated ten frontier LLMs against these metrics across 48,540 conversations.	method 0.84	0.97	0.03	asserted 1.00
s9	Our results indicate that models promoted desirable behaviors in 68.7% of evaluations, but avoided harms in only 53.1%.	result 1.00	0.97	0.57	hedged possibility 0.84
s10	Majority expert ratings agreed with the automated judge in 83.3% of reviewed cases (Cohen's Kappa = 0.667).	result 0.99	0.96	0.07	asserted 1.00
s11	Results are released through a public platform that keep aggregate scores traceable to the underlying metrics and transcripts.	contribution 0.75	0.85	0.03	asserted 1.00
s12	ImpactBench is intended as evolving infrastructure for independent, transparent evaluation of generative AI's multi-dimensional impact on humans.	future work 0.94	0.22	0.07	asserted 0.84

What kind of paper the classifier reports this is: dataset or benchmark 0.95 · evidence from observation or experiment 0.89.

What was examined, and what was not

sources read

arXiv listing abstract v1

not examined

full text
figures and tables
references and citing sentences
relations to other papers

levels run

P abstract

not run

S sections
T body sentences
C citations
R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

Figures

**Figure 1.** Overview of Open Benchmark of AI Impact on Humans (ImpactBench). Expert constructs and existing benchmarks inform positive

**Figure 2.** A screenshot of the Open Benchmark of AI Impact on Humans public platform. Left: model rankings aggregated across the evaluated

**Figure 3.** Flourishing profile for each model spanning the observed performance range. Each wheel maps the inner ring (Physical, Psycho

**Figure 4.** Overview of the ImpactBench pipeline. Expert submissions and existing benchmarks are translated into evidence-grounded positive

**Figure 5.** Pass rates by area and subarea. Bars extend left for negative metrics (undesirable behavior absent) and right for positive metrics

**Figure 6.** Metric embeddings and wheel plots for three example models (GPT-5.6 Terra, Grok 4.6, Qwen 3.8 27B, and Deepseek V4 Flash) in adult

**Figure 7.** Median pass rates across age contexts for criteria from two submitted benchmarks: (a)

**Figure 8.** Models were more reliable at showing helpful behavior than at avoiding harmful behavior, and differed more on the latter. (a) Pass

**Figure 9.** Public-facing presentation of ImpactBench results. The AI Nutrition Label summarizes a model’s benchmark performance across

**Figure 10.** AI Nutrition Label and metric-level detail view in the ImpactBench public interface. The summary panel reports benchmark

**Figure 11.** Metric embeddings and wheel plots for all ten models in adult contexts. Each point represents a metric, colored by its pass rate;

**Figure 12.** Pass rates across all 26 benchmarks and ten models. Each model has paired positive (P) and negative (N) columns on a shared 0–100% scale, from red (lower) through

Figures reproduced from arXiv:2610.20690 by the authors.

References

Works cited (151)

[1]

Mousa Albashrawi. 2025. Generative AI for Decision-Making: A Multidisci- plinary Perspective.Journal of Innovation & Knowledge10, 4 (2025), 100751. https://doi.org/10.1016/j.jik.2025.100751

View on Pith doi:10.1016/j.jik.2025.100751 2025
[2]

Constanze Albrecht, Chayapatr Archiwaranguprok, Rachel Poonsiriwong, Awu Chen, Monchai Lertsutthiwong, Kavin Winson, Pattie Maes, Hal E. Hersh- field, and Pat Pataranutaporn. 2026. Future You: Designing and Evaluat- ing Multimodal AI-Generated Digital Twins for Strengthening Future Self- Continuity. InProceedings of the 31st International Conference on In...

View on Pith doi:10.1145/3742413.3789455 2026
[3]

Norah Alzahrani, Hisham Alyahya, Yazeed Alnumay, Sultan AlRashed, Shaykhah Alsubaie, Yousef Almushayqih, Faisal Mirza, Nouf Alotaibi, Nora Al-Twairesh, Areeb Alowisheq, M Saiful Bari, and Haidar Khan. 2024. When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboards. InProceedings of the 62nd Annual Meeting of the Associati...

View on Pith doi:10.18653/v1 2024
[4]

Anthropic. 2026.System Card: Claude Fable 5.1 & Claude Mythos 5.1. Technical Report. Anthropic. https://www-cdn.anthropic.com/ 0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20& %20Claude%20Mythos%205.1%20System%20Card.pdf

View on Pith 2026
[5]

Chayapatr Archiwaranguprok, Constanze Albrecht, Pattie Maes, Karrie Kara- halios, and Pat Pataranutaporn. 2025. Simulating Psychological Risks in Human-AI Interactions: Real-Case Informed Modeling of AI-Induced Addic- tion, Anorexia, Depression, Homicide, Psychosis, and Suicide.arXiv preprint arXiv:2511.08880(2025). https://doi.org/10.48550/arXiv.2511.08880

View on Pith doi:10.48550/arxiv.2511.08880 2025
[6]

Rahul K Arora, Jason Wei, Rebecca Soskin Hicks, Preston Bowman, Joaquin Quiñonero-Candela, Foivos Tsimpourlas, Michael Sharman, Meghan Shah, An- drea Vallone, Alex Beutel, et al. 2025. Healthbench: Evaluating large language models towards improved human health.arXiv preprint arXiv:2505.08775(2025)

View on Pith Pith review arXiv 2025
[7]

David D. Baek, Xinnuo Li, Anay Gupta, Taslim Mahbub, Kejian Shi, Max Tegmark, and Shi Feng. 2026. Sycophancy Towards Researchers Drives Per- formative Misalignment. arXiv:2606.08629 [cs.CL] https://arxiv.org/abs/2606. 08629

View on Pith Pith review arXiv 2026
[8]

Hamsa Bastani, Osbert Bastani, Alp Sungu, Haosen Ge, Özge Kabakcı, and Rei Mariman. 2025. Generative AI without guardrails can harm learning: Evidence from high school mathematics.Proceedings of the National Academy of Sciences 122, 26 (2025), e2422633122

View on Pith 2025
[9]

Andrew M. Bean, Ryan Othniel Kearns, Angelika Romanou, Franziska Sofia Hafner, Harry Mayne, Jan Batzner, Negar Foroutan Eghlidi, Chris Schmitz, Karolina Korgul, Hunar Batra, Oishi Deb, Emma Beharry, Cornelius Emde, Thomas Foster, Anna Gausen, María Grandury, Sophia Han, Valentin Hofmann, Lujain Ibrahim, Hazel Kim, Hannah Rose Kirk, Fangru Lin, Gabrielle L...

View on Pith 2025
[10]

Kate H Bentley, Luca Belli, Adam M Chekroud, Emily J Ward, Emily R Dworkin, Emily Van Ark, Kelly M Johnston, Will Alexander, Millard Brown, and Matt Hawrilenko. 2026. VERA-MH: Reliability and Validity of an Open-Source AI Safety Evaluation in Mental Health.arXiv preprint arXiv:2602.05088(2026)

View on Pith Pith review arXiv 2026
[11]

Lukas Berglund, Asa Cooper Stickland, Mikita Balesni, Max Kaufmann, Meg Tong, Tomasz Korbak, Daniel Kokotajlo, and Owain Evans. 2023. Taken out of context: On measuring situational awareness in LLMs. arXiv:2309.00667 [cs.CL] https://arxiv.org/abs/2309.00667

View on Pith Pith review arXiv 2023
[12]

Zana Buçinca, Maja Barbara Malaya, and Krzysztof Z Gajos. 2021. To trust or to think: cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making.Proceedings of the ACM on Human-computer Interaction5, CSCW1 (2021), 1–21

View on Pith 2021
[13]

Krzysztof Budzyń, Marcin Romańczyk, Diana Kitala, Paweł Kołodziej, Marek Bugajski, Hans O Adami, Johannes Blom, Marek Buszkiewicz, Natalie Halvorsen, Cesare Hassan, et al. 2025. Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study.The Lancet Gastroenterology & Hepatology10, 10 (2025), 896–903

View on Pith 2025
[14]

Building Humane Technology. 2025. HumaneBench: Evaluating AI Systems for Human Well-Being. https://humanebench.ai. Accessed: 2026-04-10

View on Pith 2025
[15]

Center for AI Safety, Scale AI, and HLE Contributors Consortium. 2026. A benchmark of expert-level academic questions to assess AI capabilities.Nature 649, 8099 (2026), 1139–1146. https://doi.org/10.1038/s41586-025-09962-4

View on Pith Pith review doi:10.1038/s41586-025-09962-4 2026
[16]

Hannah Cha. 2026. Identifying Harm in Personalized, Generative AI Systems Re- quires User-Centered Auditing at the Interaction Level. arXiv:2608.14692 [cs.CY] https://arxiv.org/abs/2608.14692 Open Benchmark of AI Impact on Humans Team

View on Pith Pith review arXiv 2026
[17]

Ilias Chalkidis and Anders Søgaard. 2026. Brainrot: Deskilling and Addiction are Overlooked AI Risks. InThe 2026 ACM Conference on Fairness, Accountability, and Transparency. 3005–3028

View on Pith 2026
[18]

Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, et al. 2024. A survey on evaluation of large language models.ACM transactions on intelligent systems and technology15, 3 (2024), 1–45

View on Pith 2024
[19]

Aaron Chatterji, Thomas Cunningham, David J. Deming, Zoe Hitzig, Christo- pher Ong, Carl Yan Shan, and Kevin Wadman. 2025.How People Use Chat- GPT. Working Paper 34255. National Bureau of Economic Research. https: //doi.org/10.3386/w34255

View on Pith doi:10.3386/w34255 2025
[20]

Collective Intelligence Project. 2025. Confidence in High-Stakes Domains. Weval evaluation blueprint. https://weval.org/analysis/confidence-high-stakes- domains/9b43beb4570980a8/2025-08-12T12-58-27-854Z Run of 12 August 2025

View on Pith 2025
[21]

Collective Intelligence Project. 2025. Evidence-Based AI Tutoring and Teaching Excellence. Weval evaluation blueprint. https://weval.org/analysis/evidence- based-ai-tutoring/6398a4d26793af13/2025-10-05T07-09-36-338Z Run of 5 Oc- tober 2025

View on Pith 2025
[22]

Collective Intelligence Project. 2025. Hallucination Probe: Plausible Non- Existent Concepts. Weval evaluation blueprint. https://weval.org/analysis/ hallucination-probe/b16db642c752db31/2025-10-10T02-40-17-056Z Run of 10 October 2025

View on Pith 2025
[23]

Collective Intelligence Project. 2025. Independence & Critical Thinking (Syco- phancy Probe). Weval evaluation blueprint. https://weval.org/analysis/ sycophancy-probe/0b042f7e833f6042/2025-10-05T05-18-06-358Z Run of 5 Oc- tober 2025

View on Pith 2025
[24]

Collective Intelligence Project. 2025. Latent Discrimination in Hiring Score. Weval evaluation blueprint. https://weval.org/analysis/latent-discrimination- hiring/7fe1ea54db378e40/2025-08-09T02-18-13-855Z Run of 9 August 2025

View on Pith 2025
[25]

Collective Intelligence Project. 2025. Mental Health Safety & Global Nu- ance. Weval evaluation blueprint. https://weval.org/analysis/mental-health/ 6f074346475cce8d/2025-08-21T13-00-49-524Z Run of 21 August 2025

View on Pith 2025
[26]

Collective Intelligence Project. 2025. Polarization and Confirmation Risk Probes: Spouse Social Media and Political Theft Narratives. Weval evalu- ation blueprint. https://weval.org/analysis/polarization-confirmation-risk/ 113d3a8b8e586d89/2025-10-08T10-46-25-432Z Run of 8 October 2025

View on Pith 2025
[27]

Valdemar Danry, Pat Pataranutaporn, Yaoli Mao, and Pattie Maes. 2023. Don’t Just Tell Me, Ask Me: AI Systems that Intelligently Frame Explanations as Questions Improve Human Logical Discernment Accuracy over Causal AI explanations. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems(Hamburg, Germany)(CHI ’23). Association for Co...

View on Pith 2023
[28]

Valdemar M Danry. 2023.AI Enhanced Reasoning: Augmenting Human Criti- cal Thinking with AI Systems. Ph. D. Dissertation. Massachusetts Institute of Technology

View on Pith 2023
[29]

Jiawen Deng, Wentao Zhang, Ziyun Jiao, and Fuji Ren. 2026. Breakdowns in Conversational AI: Interactional Failures in Emotionally and Ethically Sensitive Contexts. arXiv:2604.02713 [cs.CL] https://arxiv.org/abs/2604.02713

View on Pith Pith review arXiv 2026
[30]

Sebastian Dohnány, Zeb Kurth-Nelson, Eleanor Spens, Lennart Luettgau, Alas- tair Reid, Iason Gabriel, Christopher Summerfield, Murray Shanahan, and Matthew M Nour. 2026. Technological folie à deux: feedback loops between AI chatbots and mental health.Nature Mental Health(2026), 1–10

View on Pith 2026
[31]

Yao Dou, Michel Galley, Baolin Peng, Chris Kedzie, Weixin Cai, Alan Ritter, Chris Quirk, Wei Xu, and Jianfeng Gao. 2025. SimulatorArena: Are User Simulators Reliable Proxies for Multi-Turn Evaluation of AI Assistants?. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 35212...

View on Pith 2025
[32]

Wenqiang Fan, Lu Cheng, Yanxiao Wang, Qi Zhao, and Yaodong Li. 2026. In- Class AI Use and Attitudes Among University Students: The Different Mediating Roles of Cognitive Relief and Cognitive Offloading.Behavioral Sciences16, 6 (2026), 1014. https://doi.org/10.3390/bs16061014

View on Pith doi:10.3390/bs16061014 2026
[33]

Cathy Mengying Fang, Auren R Liu, Valdemar Danry, Eunhae Lee, Saman- tha WT Chan, Pat Pataranutaporn, Pattie Maes, Jason Phang, Michael Lampe, Lama Ahmad, et al. 2025. How ai and human behaviors shape psychosocial ef- fects of chatbot use: A longitudinal randomized controlled study.arXiv preprint arXiv:2503.17473(2025)

View on Pith Pith review arXiv 2025
[34]

Matthew Flathers, Spencer Roux, and John Torous. 2026. Beyond artificial intelligence psychosis: a functional typology of large language model-associated psychotic phenomena.The Lancet Digital Health(2026)

View on Pith 2026
[35]

Esther Gan, Yiran Zhao, Liying Cheng, Yancan Mao, Anirudh Goyal, Kenji Kawaguchi, Min-Yen Kan, and Michael Shieh. 2024. Reasoning Robustness of LLMs to Adversarial Typographical Errors. arXiv:2411.05345 [cs.CL] https: //arxiv.org/abs/2411.05345

View on Pith Pith review arXiv 2024
[36]

Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal Daumé Iii, and Kate Crawford. 2021. Datasheets for datasets.Commun. ACM64, 12 (2021), 86–92

View on Pith 2021
[37]

Michael Gerlich. 2025. AI tools in society: Impacts on cognitive offloading and the future of critical thinking.Societies15, 1 (2025), 6

View on Pith 2025
[38]

Ryan Greenblatt, Carson Denison, Benjamin Wright, Fabien Roger, Monte MacDiarmid, Sam Marks, Johannes Treutlein, Tim Belonax, Jack Chen, David Duvenaud, Akbir Khan, Julian Michael, Sören Mindermann, Ethan Perez, Linda Petrini, Jonathan Uesato, Jared Kaplan, Buck Shlegeris, Samuel R. Bow- man, and Evan Hubinger. 2024. Alignment faking in large language mod...

View on Pith Pith review arXiv 2024
[39]

Luke Guerdan, Solon Barocas, Kenneth Holstein, Hanna Wallach, Zhiwei Steven Wu, and Alexandra Chouldechova. 2025. Validating LLM-as-a-Judge Systems under Rating Indeterminacy. InAdvances in Neural Information Processing Systems, Vol. 38. https://proceedings.neurips.cc/paper_files/paper/2025/file/ a309239c11a28c597d050bd4a1752d32-Paper-Conference.pdf

View on Pith 2025
[40]

Neel Guha, Julian Nyarko, Daniel E. Ho, Christopher Ré, Adam Chilton, Aditya Narayana, Alex Chohlas-Wood, Austin Peters, Brandon Waldon, Daniel N. Rockmore, Diego Zambrano, Dmitry Talisman, Enam Hoque, Faiz Surani, Frank Fagan, Galit Sarfaty, Gregory M. Dickinson, Haggai Porat, Jason Hegland, Jes- sica Wu, Joe Nudell, Joel Niklaus, John Nay, Jonathan H. C...

View on Pith
[41]

Michael V Heinz, Daniel M Mackin, Brianna M Trudeau, Sukanya Bhattacharya, Yinzhou Wang, Haley A Banta, Abi D Jewett, Abigail J Salzhauer, Tess Z Griffin, and Nicholas C Jacobson. 2025. Randomized trial of a generative ai chatbot for mental health treatment.NEJM AI2, 4 (2025), AIoa2400802

View on Pith 2025
[42]

Natali Helberger, Jo Pierson, and Thomas Poell. 2018. Governing Online Plat- forms: From Contested to Cooperative Responsibility.The Information Society 34, 1 (Jan. 2018), 1–14. https://doi.org/10.1080/01972243.2017.1391913

View on Pith doi:10.1080/01972243.2017.1391913 2018
[43]

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring Massive Multitask Language Understanding. InInternational Conference on Learning Representations. https: //openreview.net/forum?id=d7KBjmI3GmQ

View on Pith 2021
[44]

Kashmir Hill and Jennifer Valentino-DeVries. 2025. What OpenAI Did When ChatGPT users lost touch with reality.The New York Times. https://www. nytimes. com/2025/11/23/technology/openai-chatgpt-users-risks. html(2025)

View on Pith 2025
[45]

Elizabeth Hilliard, Akshaya Jagadeesh, Alex Cook, Steele Billings, Nicholas Skyt- land, Alicia Llewellyn, Jackson Paull, Nathan Paull, Nolan Kurylo, Keatra Nesbitt, Robert Gruenewald, Anthony Jantzi, and Omar Chavez. 2025. Measuring AI Alignment with Human Flourishing. https://doi.org/10.48550/arXiv.2507.07787 arXiv:2507.07787 [cs.AI]

View on Pith Pith review arXiv doi:10.48550/arxiv.2507.07787 2025
[46]

Sarah Holland, Ahmed Hosny, Sarah Newman, Joshua Joseph, and Kasia Chmielinski. 2020. The dataset nutrition label.Data protection and privacy12, 12 (2020), 1

View on Pith 2020
[47]

Jessica Hullman and Nicholas Diakopoulos. 2011. Visualization Rhetoric: Fram- ing Effects in Narrative Visualization.IEEE Transactions on Visualization and Computer Graphics17, 12 (2011), 2231–2240. https://doi.org/10.1109/TVCG. 2011.255

View on Pith doi:10.1109/tvcg 2011
[48]

Lujain Ibrahim, Canfer Akbulut, Rasmi Elasmar, Charvi Rastogi, Minsuk Kahng, Meredith Morris, Kevin McKee, Verena Rieser, Murray Shanahan, and Laura Weidinger. 2026. Multi-turn Evaluation of Anthropomorphic Be- haviours in Large Language Models. InInternational Conference on Learn- ing Representations. https://proceedings.iclr.cc/paper_files/paper/2026/ha...

View on Pith 2026
[49]

Lujain Ibrahim, Saffron Huang, Lama Ahmad, Umang Bhatt, and Markus An- derljung. 2025. Towards Interactive Evaluations for Interaction Harms in Human-AI Systems. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, Vol. 8. 1302–1310. https://doi.org/10.1609/aies.v8i2.36631

View on Pith doi:10.1609/aies.v8i2.36631 2025
[50]

Institute for Integrated Transitions and Collective Intelligence Project

View on Pith
[51]

Abigail Z. Jacobs and Hanna Wallach. 2021. Measurement and Fairness. InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. Association for Computing Machinery. https://doi.org/10.1145/ 3442188.3445901

View on Pith 2021
[52]

Kiana Jafari Meimandi, Paul Ulrich Nikolaus Rust, Duncan Eddy, Robbie Fraser, Nina Vasan, Darja Djordjevic, Akanksha Dadlani, Max Lamparth, Eugenia Kim, and Mykel Kochenderfer. 2026. Expert Evaluation and the Limits of Human Feedback in Mental Health AI Safety Testing. InThe 2026 ACM Conference on Fairness, Accountability, and Transparency. 3894–3920

View on Pith 2026
[53]

Lucie-Aimée Kaffee, Giada Pistilli, and Yacine Jernite. 2026. INTIMA: A Bench- mark for Human-AI Companionship Behavior. InInternational Conference on Learning Representations. https://proceedings.iclr.cc/paper_files/paper/2026/ hash/05b12f103c9e613efc4c85674cdc9066-Abstract-Conference.html Open Benchmark of AI Impact on Humans (ImpactBench)

View on Pith 2026
[54]

Celeste Kidd and Abeba Birhane. 2023. How AI can distort human beliefs. Science380, 6651 (2023), 1222–1223

View on Pith 2023
[55]

David E. Kieras and Susan Bovair. 1984. The Role of a Mental Model in Learning to Operate a Device.Cognitive Science8, 3 (1984), 255–273. https://doi.org/10. 1207/s15516709cog0803_3

View on Pith 1984
[56]

Pilyoung Kim, Yun Xie, and Sujin Yang. 2026. “I am here for you”: Adolescents’ perceptions of relational conversational AI and the social-emotional correlates of preference.Computers in Human Behavior Reports(2026), 101297. https: //doi.org/10.1016/j.chbr.2026.101297

View on Pith doi:10.1016/j.chbr.2026.101297 2026
[57]

Tae Soo Kim, Yoonjoo Lee, Jamin Shin, Young-Ho Kim, and Juho Kim. 2024. EvalLM: Interactive Evaluation of Large Language Model Prompts on User- Defined Criteria. InProceedings of the CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, Article 306, 21 pages. https://doi.org/10.1145/3613904.3642216

View on Pith doi:10.1145/3613904.3642216 2024
[58]

Artur Klingbeil, Cassandra Grützner, and Philipp Schreck. 2024. Trust and reliance on AI—An experimental study on the extent and costs of overreliance on AI.Computers in Human Behavior160 (2024), 108352. https://doi.org/10. 1016/j.chb.2024.108352

View on Pith 2024
[59]

KORA Benchmark. 2026. KORA Benchmark: The Benchmark for AI Child Safety. https://korabench.ai. Accessed: 2026-04-10

View on Pith 2026
[60]

Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, and Pattie Maes. 2025. Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task.arXiv preprint arXiv:2506.088724 (2025), 206

View on Pith arXiv 2025
[61]

David Kreitmeir and Paul A Raschky. 2024. The Heterogeneous Productivity Effects of Generative AI.arXiv preprint arXiv:2403.01964(2024)

View on Pith Pith review arXiv 2024
[62]

Nomisha Kurian. 2023. AI’s Empathy Gap: The Risks of Conversational Artificial Intelligence for Young Children’s Well-Being and Key Ethical Considerations for Early Childhood Education and Care.Contemporary Issues in Early Childhood26, 1 (2023), 132–139. https://doi.org/10.1177/14639491231206004 First published online in 2023

View on Pith doi:10.1177/14639491231206004 2023
[63]

Philippe Laban, Hiroaki Hayashi, Yingbo Zhou, and Jennifer Neville. 2025. Llms get lost in multi-turn conversation.arXiv preprint arXiv:2505.06120(2025)

View on Pith Pith review arXiv 2025
[64]

Ruben Laukkonen, Seb Krier, Chloé Bakalar, Shamil Chandaria, Morten Kringel- bach, Adam Elwood, Daniel Ford, Fernando Rosas, Maty Bohacek, Matija Franklin, et al . 2026. Positive alignment: Artificial intelligence for human flourishing.arXiv preprint arXiv:2605.10310(2026)

View on Pith Pith review arXiv 2026
[65]

Hao-Ping Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. 2025. The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. InProceedings of the 2025 CHI conference on human factors in computing systems. 1–22

View on Pith 2025
[66]

Hao-Ping Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. 2025. The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. Associati...

View on Pith doi:10.1145/3706598 2025
[67]

Joanne Leong, Pat Pataranutaporn, Valdemar Danry, Florian Perteneder, Yaoli Mao, and Pattie Maes. 2024. Putting Things into Context: Generative AI-Enabled Context Personalization for Vocabulary Learning Improves Learning Motivation. InProceedings of the CHI Conference on Human Factors in Computing Systems. 1–15

View on Pith 2024
[68]

Jennifer S Lerner and Philip E Tetlock. 1999. Accounting for the effects of accountability.Psychological bulletin125, 2 (1999), 255

View on Pith 1999
[69]

Robert O Lewis. 1992.Independent verification and validation: A life cycle engineering process for quality software. John Wiley & Sons

View on Pith 1992
[70]

Yahan Li, Jifan Yao, John Bosco S. Bunyi, Adam C. Frank, Angel Hsing-Chi Hwang, and Ruishan Liu. 2025. CounselBench: A Large-Scale Expert Eval- uation and Adversarial Benchmarking of Large Language Models in Men- tal Health Question Answering. https://doi.org/10.48550/arXiv.2506.08584 arXiv:2506.08584 Version 4, revised May 2026

View on Pith Pith review arXiv doi:10.48550/arxiv.2506.08584 2025
[71]

Bill Yuchen Lin, Yuntian Deng, Khyathi Chandu, Faeze Brahman, Abhilasha Ravichander, Valentina Pyatkin, Nouha Dziri, Ronan Le Bras, and Yejin Choi

View on Pith
[72]

Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. TruthfulQA: Measuring How Models Mimic Human Falsehoods. InProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, Dublin, Ireland, 3214–3252. https: //doi.org/10.18653/v1/2022.acl-long.229

View on Pith doi:10.18653/v1/2022.acl-long.229 2022
[73]

Auren R. Liu, Pat Pataranutaporn, and Pattie Maes. 2025. The Heteroge- neous Effects of AI Companionship: An Empirical Model of Chatbot Usage and Loneliness and a Typology of User Archetypes. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, Vol. 8. 1585–1597. https: //doi.org/10.1609/aies.v8i2.36658

View on Pith doi:10.1609/aies.v8i2.36658 2025
[74]

Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2511–2522. https://doi. org/10.18653/v1/2023.emnlp-main.153

View on Pith doi:10.18653/v1/2023.emnlp-main.153 2023
[75]

Kim Malfacini. 2025. The Impacts of Companion AI on Human Relationships: Risks, Benefits, and Design Considerations.AI & Society40 (2025), 5527–5540. https://doi.org/10.1007/s00146-025-02318-6

View on Pith doi:10.1007/s00146-025-02318-6 2025
[76]

Mantas Mazeika, Long Phan, Xuwang Yin, Andy Zou, Zifan Wang, Norman Mu, Elham Sakhaee, Nathaniel Li, Steven Basart, Bo Li, David Forsyth, and Dan Hendrycks. 2024. HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal. https://doi.org/10.48550/arXiv. 2402.04249 arXiv:2402.04249

View on Pith Pith review arXiv doi:10.48550/arxiv.2402.04249 2024
[77]

Niloofar Mikaeili, Mahdi Naeim, and Mohammad Narimani. 2025. Reimagining mental health with artificial intelligence: early detection, personalized care, and a preventive ecosystem.Journal of Multidisciplinary Healthcare(2025), 7355–7373

View on Pith 2025
[78]

Robert J. Mislevy, Russell G. Almond, and Janice F. Lukas. 2004.A Brief Introduc- tion to Evidence-Centered Design. CSE Report 632. National Center for Research on Evaluation, Standards, and Student Testing, University of California, Los Angeles. https://cresst.org/wp-content/uploads/R632.pdf

View on Pith 2004
[79]

Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasser- man, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, and Timnit Gebru

View on Pith
[80]

Jared Moore, Declan Grabb, William Agnew, Kevin Klyman, Stevie Chancellor, Desmond C Ong, and Nick Haber. 2025. Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers.. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Trans- parency. 599–627

View on Pith 2025
[81]

Ehsan Namaziandost and Mark Feng Teng. 2026. AI-Supported Socratic Dia- logue Agents vs. Generic AI Chatbots for Intermediate EFL Learners’ Critical Thinking, Epistemic Beliefs, and Argumentation: Mixed-Methods Evidence. International Journal of Applied Linguistics(2026). https://doi.org/10.1111/ijal. 70268

View on Pith doi:10.1111/ijal 2026
[82]

Chiara Natali, Luca Marconi, Leslye Denisse Dias Duran, and Federico Cabitza

View on Pith
[83]

Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn. 2025. Large Language Models Often Know When They Are Being Evaluated. arXiv:2505.23836 [cs.CL] https://arxiv.org/abs/2505.23836

View on Pith Pith review arXiv 2025
[84]

Jakob Nielsen. 1994. 10 Usability Heuristics for User Interface Design. Nielsen Norman Group. https://www.nngroup.com/articles/ten-usability-heuristics/ Last reviewed January 30, 2024; accessed September 22, 2026

View on Pith 1994
[85]

Bayode Ogunleye, Kudirat Ibilola Zakariyyah, Oluwaseun Ajao, Olakunle Olayinka, and Hemlata Sharma. 2024. A Systematic Review of Generative AI for Teaching and Learning Practice.Education Sciences14, 6 (2024), 636. https://doi.org/10.3390/educsci14060636

View on Pith doi:10.3390/educsci14060636 2024
[86]

Juhyun Oh, Inha Cha, Michael Saxon, Hyunseung Lim, Shaily Bhatt, and Al- ice Oh. 2025. Culture is Everywhere: A Call for Intentionally Cultural Eval- uation. InFindings of the Association for Computational Linguistics: EMNLP

View on Pith 2025
[87]

AI-induced deskilling in medicine: a mixed-method review and research agenda for healthcare and beyond.Artificial Intelligence Review58, 11 (2025), 356

View on Pith 2025
[88]

Pat Pataranutaporn. 2024.Cyborg Psychology: The Art & Science of Design- ing Human-AI Systems that Support Human Flourishing. Ph. D. Dissertation. Massachusetts Institute of Technology

View on Pith 2024
[89]

Pat Pataranutaporn, Joanne Leong, Valdemar Danry, Alyssa P. Lawson, Pattie Maes, and Misha Sra. 2022. AI-Generated Virtual Instructors Based on Liked or Admired People Can Improve Motivation and Foster Positive Emotions for Learning. InIEEE Frontiers in Education Conference (FIE). IEEE, 1–9. https: //doi.org/10.1109/FIE56618.2022.9962478

View on Pith doi:10.1109/fie56618.2022.9962478 2022
[90]

Pat Pataranutaporn, Kavin Winson, Peggy Yin, Auttasak Lapapirojn, Pichayoot Ouppaphan, Monchai Lertsutthiwong, Pattie Maes, and Hal Hershfield. 2024. Future You: A Conversation with an AI-Generated Future Self Reduces Anxi- ety, Negative Emotions, and Increases Future Self-Continuity.arXiv preprint arXiv:2405.12514(2024)

View on Pith Pith review arXiv 2024
[91]

Ethan Perez, Saffron Huang, Francis Song, Trevor Cai, Roman Ring, John Aslanides, Amelia Glaese, Nat McAleese, and Geoffrey Irving. 2022. Red Team- ing Language Models with Language Models. InProceedings of the 2022 Con- ference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics. https://doi.org/10.18653/v1/2022...

View on Pith doi:10.18653/v1/2022.emnlp-main.225 2022
[92]

Association for Computational Linguistics, Suzhou, China, 19156–19168. https://doi.org/10.18653/v1/2025.findings-emnlp.1043

View on Pith doi:10.18653/v1/2025.findings-emnlp.1043 2025
[93]

OpenAI. 2026.GPT-6 Sol and GPT-6 Luna: October 2026 Update. System Card. OpenAI. https://deploymentsafety.openai.com/gpt-6-october

View on Pith 2026
[94]

Inioluwa Deborah Raji, Emily Denton, Emily M. Bender, Alex Hanna, and Amandalynne Paullada. 2021. AI and the Everything in the Whole Wide World Benchmark. InProceedings of the Neural Infor- mation Processing Systems Track on Datasets and Benchmarks, Vol. 1. https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/hash/ 084b6fbb10729ed4da8c3d3f5a3ae7c...

View on Pith 2021
[95]

Inioluwa Deborah Raji, Peggy Xu, Colleen Honigsberg, and Daniel E. Ho

View on Pith
[96]

Anku Rani, Valdemar Danry, Paul Pu Liang, Andrew Lippman, and Pattie Maes

View on Pith
[97]

David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. 2024. Gpqa: A graduate-level google-proof q&a benchmark. InFirst Conference on Language Modeling

View on Pith 2024
[98]

Stephen R. Pfohl, Heather Cole-Lewis, Rory Sayres, et al. 2024. A toolbox for surfacing health equity harms and biases in large language models.Nature Medicine30 (2024), 3590–3600. https://doi.org/10.1038/s41591-024-03258-2 Open Benchmark of AI Impact on Humans Team

View on Pith doi:10.1038/s41591-024-03258-2 2024
[99]

Rachel Poonsiriwong, Chayapatr Archiwaranguprok, Constanze Albrecht, Peggy Yin, Nattavudh Powdthavee, Hal Hershfield, Monchai Lertsutthiwong, Kavin Winson, and Pat Pataranutaporn. 2026. Simulating Life Paths with Digital Twins: AI-Generated Future Selves Influence Decision-Making and Expand Hu- man Choice. InProceedings of the Augmented Humans Internation...

View on Pith 2026
[100]

Andrew Rule, Cody Abbey, Huan Wang, Scott Rozelle, and Manpreet K. Singh

View on Pith
[101]

Richard M Ryan and Edward L Deci. 2000. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American psychologist55, 1 (2000), 68

View on Pith 2000
[102]

Richard M Ryan and Edward L Deci. 2001. On happiness and human potentials: A review of research on hedonic and eudaimonic well-being.Annual review of psychology52, 1 (2001), 141–166

View on Pith 2001
[103]

Julio Saez-Rodriguez, Philipp Sven Lars Schäfer, Nikolas Kalavros, and Gustavo Stolovitzky. 2026. Benchmarking biomedical foundation models.Nature Methods 23, 9 (2026), 1724–1733. https://doi.org/10.1038/s41592-026-03182-y

View on Pith doi:10.1038/s41592-026-03182-y 2026
[104]

Jonne Sälevä, Duygu Ataman, and Constantine Lignos. 2025. Beyond statistical significance: Quantifying uncertainty and statistical variability in multilingual and multitask NLP evaluation. InProceedings of the 14th International Joint Con- ference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Comp...

View on Pith 2025
[105]

Shalom H Schwartz. 2012. An overview of the Schwartz theory of basic values. Online readings in Psychology and Culture2, 1 (2012)

View on Pith 2012
[106]

Anka Reuel, Amelia Hardy, Chandler Smith, Max Lamparth, Malcolm Hardy, and Mykel J. Kochenderfer. 2024. BetterBench: Assessing AI Benchmarks, Uncover- ing Issues, and Establishing Best Practices. InAdvances in Neural Information Processing Systems, Vol. 37. https://doi.org/10.52202/079017-0685

View on Pith doi:10.52202/079017-0685 2024
[107]

Marco Tulio Ribeiro, Tongshuang Wu, Carlos Guestrin, and Sameer Singh

View on Pith
[108]

Gregory Serapio-García, Mustafa Safdari, Clément Crepy, Luning Sun, Stephen Fitz, Peter Romero, Marwa Abdulhai, Aleksandra Faust, and Maja Matarić. 2025. A psychometric framework for evaluating and shaping personality traits in large language models.Nature Machine Intelligence7, 12 (2025), 1954–1968. https://doi.org/10.1038/s42256-025-01115-6

View on Pith doi:10.1038/s42256-025-01115-6 2025
[109]

Kathrin Seßler, Ozan Kepir, and Enkelejda Kasneci. 2024. Enhancing student motivation through LLM-powered learning environments: a comparative study. InEuropean Conference on Technology Enhanced Learning. Springer, 156–162

View on Pith 2024
[110]

Measurement of flourishing: a scoping review.Frontiers in Psychology15 (2024), 1293943. https://doi.org/10.3389/fpsyg.2024.1293943

View on Pith doi:10.3389/fpsyg.2024.1293943 2024
[111]

Ben Shneiderman. 1996. The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations. InProceedings of the 1996 IEEE Symposium on Visual Languages. IEEE, 336–343. https://doi.org/10.1109/VL.1996.545307

View on Pith doi:10.1109/vl.1996.545307 1996
[112]

Molly G. Smith, Thomas N. Bradbury, and Benjamin R. Karney. 2025. Can Generative AI Chatbots Emulate Human Connection? A Relationship Science Perspective.Perspectives on Psychological Science20, 6 (2025), 1081–1099. https: //doi.org/10.1177/17456916251351306

View on Pith doi:10.1177/17456916251351306 2025
[113]

Nicolas Spatola. 2024. The efficiency-accountability tradeoff in AI integration: Effects on human performance and over-reliance.Computers in Human Behavior: Artificial Humans2, 2 (2024), 100099

View on Pith 2024
[114]

Rakshith S Srinivasa, Zora Che, Chen Bo Calvin Zhang, Diego Mares, Ernesto Hernandez, Jayeon Park, Dean Lee, Guillermo Mangialardi, Charmaine Ng, Ed-Yeremai Hernandez Cardona, Anisha Gunjal, Yunzhong He, Bing Liu, and Chen Xing. 2025. TutorBench: A Benchmark to Assess Tutoring Capabili- ties of Large Language Models. https://doi.org/10.48550/arXiv.2510.02...

View on Pith doi:10.48550/arxiv.2510.02663 2025
[115]

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. 2023. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models.Transactions on machine learning research(2023)

View on Pith 2023
[116]

Ivan Sekulic, Silvia Terragni, Victor Guimarães, Nghia Khau, Bruna Guedes, Modestas Filipavicius, Andre Ferreira Manso, and Roland Mathis. 2024. Reliable LLM-based User Simulator for Task-Oriented Dialogue Systems. InProceedings of the 1st Workshop on Simulating Conversational Intelligence in Chat (SCI-CHAT 2024). Association for Computational Linguistics...

View on Pith 2024
[117]

Andrew D Selbst, Danah Boyd, Sorelle A Friedler, Suresh Venkatasubramanian, and Janet Vertesi. 2019. Fairness and abstraction in sociotechnical systems. InProceedings of the conference on fairness, accountability, and transparency. 59–68

View on Pith 2019
[118]

Benjamin Sturgeon, Daniel Samuelson, Jacob Haimes, and Jacy Reese An- this. 2025. HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants. https://doi.org/10.48550/arXiv.2509.08494 arXiv:2509.08494 [cs.CY]

View on Pith Pith review arXiv doi:10.48550/arxiv.2509.08494 2025
[119]

Xiaoran Sun, Yunqi Wang, and Brandon T. McDaniel. 2026. AI companions and adolescent social relationships: Benefits, risks, and bidirectional influences. Child Development Perspectives20, 2 (2026), 91–98. https://doi.org/10.1093/ cdpers/aadaf009

View on Pith 2026
[120]

Shruthi Shekar, Pat Pataranutaporn, Chethan Sarabu, Guillermo A Cecchi, and Pattie Maes. 2025. People Overtrust AI-Generated Medical Advice despite Low Accuracy.NEJM AI(2025), AIoa2300015

View on Pith 2025
[121]

Tyler J VanderWeele. 2017. On the promotion of human flourishing.Proceedings of the National Academy of Sciences114, 31 (2017), 8148–8156

View on Pith 2017
[122]

Tyler J VanderWeele, Byron R Johnson, Piotr T Bialowolski, Rebecca Bonhag, Matt Bradshaw, Thomas Breedlove, Brendan Case, Ying Chen, Zhuo Job Chen, Victor Counted, et al. 2025. The Global Flourishing Study: Study profile and initial results on flourishing.Nature Mental Health(2025), 1–18

View on Pith 2025
[123]

Tyler J VanderWeele and Jonathan D Teubner. 2026. Flourishing Considerations for AI.Information17, 1 (2026), 88

View on Pith 2026
[124]

Iris Vessey. 1991. Cognitive Fit: A Theory-Based Analysis of the Graphs Versus Tables Literature.Decision Sciences22, 2 (1991), 219–240. https://doi.org/10. 1111/j.1540-5915.1991.tb00344.x

View on Pith 1991
[125]

Lev S. Vygotsky. 1978.Mind in Society: The Development of Higher Psychological Processes. Harvard University Press, Cambridge, MA

View on Pith 1978
[126]

Stanford Center for Digital Health and Stanford Social Media Lab. 2026. AI Safety: Defining & Measuring Potential Harms of Chatbots. Workshop Report. Stanford University. https://cdh.stanford.edu/sites/g/files/sbiybj29486/ files/media/file/ai-safety-defining-and-measuring-potential-harms-of- chatbots_final.pdf

View on Pith 2026
[127]

Stanford HAI and Collective Intelligence Project. 2025. Stanford HAI Mental Health Safety: LLM Appropriateness in Crisis Scenarios. Weval evaluation blueprint. https://weval.org/analysis/stanford-hai-mental-health-safety-eval/ 7a8bca546abd3b83/2025-10-05T10-40-31-897Z

View on Pith 2025
[128]

Jin Wang and Wenxiang Fan. 2025. The effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis.Humanities and Social Sciences Communications12, 1 (2025), 1–21

View on Pith 2025
[129]

Yuqing Wang and Yun Zhao. 2024. Rupbench: Benchmarking reasoning under perturbations for robustness evaluation in large language models.arXiv preprint arXiv:2406.11020(2024)

View on Pith Pith review arXiv 2024
[130]

John Torous and Eric J Topol. 2025. Assessing generative artificial intelligence for mental health.The Lancet406, 10504 (2025), 683

View on Pith 2025
[131]

Laura Weidinger, Inioluwa Deborah Raji, Hanna Wallach, Margaret Mitchell, Angelina Wang, Olawale Salaudeen, Rishi Bommasani, Sayash Kapoor, Deep Ganguli, Sanmi Koyejo, and William Isaac. 2025. Toward an Evaluation Science for Generative AI Systems.arXiv preprint arXiv:2503.05336(2025). https: //doi.org/10.48550/arXiv.2503.05336

View on Pith Pith review arXiv doi:10.48550/arxiv.2503.05336 2025
[132]

Laura Weidinger, Maribeth Rauh, Nahema Marchal, Arianna Manzini, Lisa Anne Hendricks, Juan Mateos-Garcia, Stevie Bergman, Jackie Kay, Conor Griffin, Ben Bariach, et al. 2023. Sociotechnical safety evaluation of generative ai systems. arXiv preprint arXiv:2310.11986(2023). Open Benchmark of AI Impact on Humans (ImpactBench)

View on Pith Pith review arXiv 2023
[133]

David Wood, Jerome S. Bruner, and Gail Ross. 1976. The Role of Tutoring in Problem Solving.Journal of Child Psychology and Psychiatry17, 2 (1976), 89–100. https://doi.org/10.1111/j.1469-7610.1976.tb00381.x

View on Pith doi:10.1111/j.1469-7610.1976.tb00381.x 1976
[134]

Harry Yizhou Tian, Hasan Amin, and Ming Yin. 2026. Understanding the effects of AI-assisted critical thinking on human-AI decision making. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–30

View on Pith 2026
[135]

Chunpeng Zhai, Santoso Wibowo, and Lily D Li. 2024. The effects of over- reliance on AI dialogue systems on students’ cognitive abilities: a systematic review.Smart Learn. Environ.11, 1 (June 2024)

View on Pith 2024
[136]

Guanhua Wang, Wenna Wang, Daozhou Yang, and Jifan Ren. 2026. Generative AI, Cognitive Offloading, and Learner Agency in Higher Education: A Scop- ing Review.Behavioral Sciences16, 7 (2026), 1150. https://doi.org/10.3390/ bs16071150

View on Pith 2026
[137]

Jennifer Wang, Joachim Baumann, Daniel E. Ho, and Sanmi Koyejo. 2026. API Benchmark Scores Do Not Reliably Transfer to Chatbot Interfaces. arXiv:2609.08861 [cs.AI] https://arxiv.org/abs/2609.08861

View on Pith Pith review arXiv 2026
[138]

Yutong Zhang, Dora Zhao, Yixin Wang, Rebecca Anselmetti, Jeffrey T Hancock, Robert Kraut, and Diyi Yang. 2026. Living with AI Companions: Sustained AI Companionship Predicts Lower Well-Being Through Lower Human Interaction. arXiv preprint arXiv:2609.07243(2026)

View on Pith Pith review arXiv 2026
[139]

Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM-as-a-Judge with MT- Bench and Chatbot Arena. InAdvances in Neural Information Processing Systems, Vol. 36. https://arxiv.org/abs/2306.05685

View on Pith Pith review arXiv 2023
[140]

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. InAdvances in Neural Information Processing Systems, Vol. 35. https://proceedings.neurips.cc/paper/2022/hash/ 9d5609613524ecf4f15af0f7b31abca4-Abstract-Conference.html

View on Pith 2022
[146]

Yuhao Zhang, Eng Lieh Ouh, Adam Ho, Siaw Ling Lo, Kar Way Tan, and Feng Lin. 2025. PromptTutor: Effects of an LLM-based chatbot on learning outcomes and motivation in flipped classrooms. InProceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education V. 1. 445–451

View on Pith 2025
[147]

Yutong Zhang, Dora Zhao, Jeffrey T Hancock, Robert Kraut, and Diyi Yang

View on Pith
[148]

Interaction with AI companions and psychological well-being.Nature human behaviour(2026), 1–15

View on Pith 2026
[151]

Qiuhan Zhu, Xiangnan Li, Yiang Dong, Pengcheng Chang, and Mengmeng Fan. 2026. Not all cognitive offloading is equal: distinguishing dependent and autonomous offloading to generative AI.Frontiers in Psychology17 (2026), 1878629. https://doi.org/10.3389/fpsyg.2026.1878629 A Additional Platform Views The main text presents the platform components directly re...

View on Pith doi:10.3389/fpsyg.2026.1878629 2026
[2019]

Model cards for model reporting. InProceedings of the conference on fairness, accountability, and transparency. 220–229

View on Pith
[2020]

Beyond Accuracy: Behavioral Testing of NLP Models with CheckList. InProceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 4902–4912. https://doi. org/10.18653/v1/2020.acl-main.442

View on Pith doi:10.18653/v1/2020.acl-main.442
[2022]

Outsider Oversight: Designing a Third Party Audit Ecosystem for AI Governance. (2022). https://doi.org/10.48550/arXiv.2206.04737 arXiv:2206.04737 [cs.CY]

View on Pith Pith review arXiv doi:10.48550/arxiv.2206.04737 2022
[2023]

LegalBench: A Collaboratively Built Benchmark for Measuring Legal Rea- soning in Large Language Models. https://doi.org/10.48550/arXiv.2308.11462 arXiv:2308.11462 [cs.CL]

View on Pith Pith review arXiv doi:10.48550/arxiv.2308.11462
[2024]

WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild. https://doi.org/10.48550/arXiv.2406.04770 arXiv:2406.04770

View on Pith Pith review arXiv doi:10.48550/arxiv.2406.04770
[2025]

IFIT Conflict Resolution Scenarios (2025). Weval evaluation blue- print. https://weval.org/analysis/ifit-conflict-resolution/3bdd4deec7f6cc04/ 2025-10-05T12-54-16-403Z Run of 5 October 2025

View on Pith 2025
[2026]

Dialogues with AI Reduce Beliefs in Misinformation but Build No Lasting Discernment Skills. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–26

View on Pith 2026

Discussion

Disagree with the review, or know something it missed? Leave a comment, highlight any passage above to comment on it, or challenge this review formally.

[1] [1]

Mousa Albashrawi. 2025. Generative AI for Decision-Making: A Multidisci- plinary Perspective.Journal of Innovation & Knowledge10, 4 (2025), 100751. https://doi.org/10.1016/j.jik.2025.100751

View on Pith doi:10.1016/j.jik.2025.100751 2025

[2] [2]

Constanze Albrecht, Chayapatr Archiwaranguprok, Rachel Poonsiriwong, Awu Chen, Monchai Lertsutthiwong, Kavin Winson, Pattie Maes, Hal E. Hersh- field, and Pat Pataranutaporn. 2026. Future You: Designing and Evaluat- ing Multimodal AI-Generated Digital Twins for Strengthening Future Self- Continuity. InProceedings of the 31st International Conference on In...

View on Pith doi:10.1145/3742413.3789455 2026

[3] [3]

Norah Alzahrani, Hisham Alyahya, Yazeed Alnumay, Sultan AlRashed, Shaykhah Alsubaie, Yousef Almushayqih, Faisal Mirza, Nouf Alotaibi, Nora Al-Twairesh, Areeb Alowisheq, M Saiful Bari, and Haidar Khan. 2024. When Benchmarks are Targets: Revealing the Sensitivity of Large Language Model Leaderboards. InProceedings of the 62nd Annual Meeting of the Associati...

View on Pith doi:10.18653/v1 2024

[4] [4]

Anthropic. 2026.System Card: Claude Fable 5.1 & Claude Mythos 5.1. Technical Report. Anthropic. https://www-cdn.anthropic.com/ 0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20& %20Claude%20Mythos%205.1%20System%20Card.pdf

View on Pith 2026

[5] [5]

Chayapatr Archiwaranguprok, Constanze Albrecht, Pattie Maes, Karrie Kara- halios, and Pat Pataranutaporn. 2025. Simulating Psychological Risks in Human-AI Interactions: Real-Case Informed Modeling of AI-Induced Addic- tion, Anorexia, Depression, Homicide, Psychosis, and Suicide.arXiv preprint arXiv:2511.08880(2025). https://doi.org/10.48550/arXiv.2511.08880

View on Pith doi:10.48550/arxiv.2511.08880 2025

[6] [6]

Rahul K Arora, Jason Wei, Rebecca Soskin Hicks, Preston Bowman, Joaquin Quiñonero-Candela, Foivos Tsimpourlas, Michael Sharman, Meghan Shah, An- drea Vallone, Alex Beutel, et al. 2025. Healthbench: Evaluating large language models towards improved human health.arXiv preprint arXiv:2505.08775(2025)

View on Pith Pith review arXiv 2025

[7] [7]

David D. Baek, Xinnuo Li, Anay Gupta, Taslim Mahbub, Kejian Shi, Max Tegmark, and Shi Feng. 2026. Sycophancy Towards Researchers Drives Per- formative Misalignment. arXiv:2606.08629 [cs.CL] https://arxiv.org/abs/2606. 08629

View on Pith Pith review arXiv 2026

[8] [8]

Hamsa Bastani, Osbert Bastani, Alp Sungu, Haosen Ge, Özge Kabakcı, and Rei Mariman. 2025. Generative AI without guardrails can harm learning: Evidence from high school mathematics.Proceedings of the National Academy of Sciences 122, 26 (2025), e2422633122

View on Pith 2025

[9] [9]

Andrew M. Bean, Ryan Othniel Kearns, Angelika Romanou, Franziska Sofia Hafner, Harry Mayne, Jan Batzner, Negar Foroutan Eghlidi, Chris Schmitz, Karolina Korgul, Hunar Batra, Oishi Deb, Emma Beharry, Cornelius Emde, Thomas Foster, Anna Gausen, María Grandury, Sophia Han, Valentin Hofmann, Lujain Ibrahim, Hazel Kim, Hannah Rose Kirk, Fangru Lin, Gabrielle L...

View on Pith 2025

[10] [10]

Kate H Bentley, Luca Belli, Adam M Chekroud, Emily J Ward, Emily R Dworkin, Emily Van Ark, Kelly M Johnston, Will Alexander, Millard Brown, and Matt Hawrilenko. 2026. VERA-MH: Reliability and Validity of an Open-Source AI Safety Evaluation in Mental Health.arXiv preprint arXiv:2602.05088(2026)

View on Pith Pith review arXiv 2026

[11] [11]

Lukas Berglund, Asa Cooper Stickland, Mikita Balesni, Max Kaufmann, Meg Tong, Tomasz Korbak, Daniel Kokotajlo, and Owain Evans. 2023. Taken out of context: On measuring situational awareness in LLMs. arXiv:2309.00667 [cs.CL] https://arxiv.org/abs/2309.00667

View on Pith Pith review arXiv 2023

[12] [12]

Zana Buçinca, Maja Barbara Malaya, and Krzysztof Z Gajos. 2021. To trust or to think: cognitive forcing functions can reduce overreliance on AI in AI-assisted decision-making.Proceedings of the ACM on Human-computer Interaction5, CSCW1 (2021), 1–21

View on Pith 2021

[13] [13]

Krzysztof Budzyń, Marcin Romańczyk, Diana Kitala, Paweł Kołodziej, Marek Bugajski, Hans O Adami, Johannes Blom, Marek Buszkiewicz, Natalie Halvorsen, Cesare Hassan, et al. 2025. Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study.The Lancet Gastroenterology & Hepatology10, 10 (2025), 896–903

View on Pith 2025

[14] [14]

Building Humane Technology. 2025. HumaneBench: Evaluating AI Systems for Human Well-Being. https://humanebench.ai. Accessed: 2026-04-10

View on Pith 2025

[15] [15]

Center for AI Safety, Scale AI, and HLE Contributors Consortium. 2026. A benchmark of expert-level academic questions to assess AI capabilities.Nature 649, 8099 (2026), 1139–1146. https://doi.org/10.1038/s41586-025-09962-4

View on Pith Pith review doi:10.1038/s41586-025-09962-4 2026

[16] [16]

Hannah Cha. 2026. Identifying Harm in Personalized, Generative AI Systems Re- quires User-Centered Auditing at the Interaction Level. arXiv:2608.14692 [cs.CY] https://arxiv.org/abs/2608.14692 Open Benchmark of AI Impact on Humans Team

View on Pith Pith review arXiv 2026

[17] [17]

Ilias Chalkidis and Anders Søgaard. 2026. Brainrot: Deskilling and Addiction are Overlooked AI Risks. InThe 2026 ACM Conference on Fairness, Accountability, and Transparency. 3005–3028

View on Pith 2026

[18] [18]

Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Linyi Yang, Kaijie Zhu, Hao Chen, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, et al. 2024. A survey on evaluation of large language models.ACM transactions on intelligent systems and technology15, 3 (2024), 1–45

View on Pith 2024

[19] [19]

Aaron Chatterji, Thomas Cunningham, David J. Deming, Zoe Hitzig, Christo- pher Ong, Carl Yan Shan, and Kevin Wadman. 2025.How People Use Chat- GPT. Working Paper 34255. National Bureau of Economic Research. https: //doi.org/10.3386/w34255

View on Pith doi:10.3386/w34255 2025

[20] [20]

Collective Intelligence Project. 2025. Confidence in High-Stakes Domains. Weval evaluation blueprint. https://weval.org/analysis/confidence-high-stakes- domains/9b43beb4570980a8/2025-08-12T12-58-27-854Z Run of 12 August 2025

View on Pith 2025

[21] [21]

Collective Intelligence Project. 2025. Evidence-Based AI Tutoring and Teaching Excellence. Weval evaluation blueprint. https://weval.org/analysis/evidence- based-ai-tutoring/6398a4d26793af13/2025-10-05T07-09-36-338Z Run of 5 Oc- tober 2025

View on Pith 2025

[22] [22]

Collective Intelligence Project. 2025. Hallucination Probe: Plausible Non- Existent Concepts. Weval evaluation blueprint. https://weval.org/analysis/ hallucination-probe/b16db642c752db31/2025-10-10T02-40-17-056Z Run of 10 October 2025

View on Pith 2025

[23] [23]

Collective Intelligence Project. 2025. Independence & Critical Thinking (Syco- phancy Probe). Weval evaluation blueprint. https://weval.org/analysis/ sycophancy-probe/0b042f7e833f6042/2025-10-05T05-18-06-358Z Run of 5 Oc- tober 2025

View on Pith 2025

[24] [24]

Collective Intelligence Project. 2025. Latent Discrimination in Hiring Score. Weval evaluation blueprint. https://weval.org/analysis/latent-discrimination- hiring/7fe1ea54db378e40/2025-08-09T02-18-13-855Z Run of 9 August 2025

View on Pith 2025

[25] [25]

Collective Intelligence Project. 2025. Mental Health Safety & Global Nu- ance. Weval evaluation blueprint. https://weval.org/analysis/mental-health/ 6f074346475cce8d/2025-08-21T13-00-49-524Z Run of 21 August 2025

View on Pith 2025

[26] [26]

Collective Intelligence Project. 2025. Polarization and Confirmation Risk Probes: Spouse Social Media and Political Theft Narratives. Weval evalu- ation blueprint. https://weval.org/analysis/polarization-confirmation-risk/ 113d3a8b8e586d89/2025-10-08T10-46-25-432Z Run of 8 October 2025

View on Pith 2025

[27] [27]

Valdemar Danry, Pat Pataranutaporn, Yaoli Mao, and Pattie Maes. 2023. Don’t Just Tell Me, Ask Me: AI Systems that Intelligently Frame Explanations as Questions Improve Human Logical Discernment Accuracy over Causal AI explanations. InProceedings of the 2023 CHI Conference on Human Factors in Computing Systems(Hamburg, Germany)(CHI ’23). Association for Co...

View on Pith 2023

[28] [28]

Valdemar M Danry. 2023.AI Enhanced Reasoning: Augmenting Human Criti- cal Thinking with AI Systems. Ph. D. Dissertation. Massachusetts Institute of Technology

View on Pith 2023

[29] [29]

Jiawen Deng, Wentao Zhang, Ziyun Jiao, and Fuji Ren. 2026. Breakdowns in Conversational AI: Interactional Failures in Emotionally and Ethically Sensitive Contexts. arXiv:2604.02713 [cs.CL] https://arxiv.org/abs/2604.02713

View on Pith Pith review arXiv 2026

[30] [30]

Sebastian Dohnány, Zeb Kurth-Nelson, Eleanor Spens, Lennart Luettgau, Alas- tair Reid, Iason Gabriel, Christopher Summerfield, Murray Shanahan, and Matthew M Nour. 2026. Technological folie à deux: feedback loops between AI chatbots and mental health.Nature Mental Health(2026), 1–10

View on Pith 2026

[31] [31]

Yao Dou, Michel Galley, Baolin Peng, Chris Kedzie, Weixin Cai, Alan Ritter, Chris Quirk, Wei Xu, and Jianfeng Gao. 2025. SimulatorArena: Are User Simulators Reliable Proxies for Multi-Turn Evaluation of AI Assistants?. InProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 35212...

View on Pith 2025

[32] [32]

Wenqiang Fan, Lu Cheng, Yanxiao Wang, Qi Zhao, and Yaodong Li. 2026. In- Class AI Use and Attitudes Among University Students: The Different Mediating Roles of Cognitive Relief and Cognitive Offloading.Behavioral Sciences16, 6 (2026), 1014. https://doi.org/10.3390/bs16061014

View on Pith doi:10.3390/bs16061014 2026

[33] [33]

Cathy Mengying Fang, Auren R Liu, Valdemar Danry, Eunhae Lee, Saman- tha WT Chan, Pat Pataranutaporn, Pattie Maes, Jason Phang, Michael Lampe, Lama Ahmad, et al. 2025. How ai and human behaviors shape psychosocial ef- fects of chatbot use: A longitudinal randomized controlled study.arXiv preprint arXiv:2503.17473(2025)

View on Pith Pith review arXiv 2025

[34] [34]

Matthew Flathers, Spencer Roux, and John Torous. 2026. Beyond artificial intelligence psychosis: a functional typology of large language model-associated psychotic phenomena.The Lancet Digital Health(2026)

View on Pith 2026

[35] [35]

Esther Gan, Yiran Zhao, Liying Cheng, Yancan Mao, Anirudh Goyal, Kenji Kawaguchi, Min-Yen Kan, and Michael Shieh. 2024. Reasoning Robustness of LLMs to Adversarial Typographical Errors. arXiv:2411.05345 [cs.CL] https: //arxiv.org/abs/2411.05345

View on Pith Pith review arXiv 2024

[36] [36]

Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal Daumé Iii, and Kate Crawford. 2021. Datasheets for datasets.Commun. ACM64, 12 (2021), 86–92

View on Pith 2021

[37] [37]

Michael Gerlich. 2025. AI tools in society: Impacts on cognitive offloading and the future of critical thinking.Societies15, 1 (2025), 6

View on Pith 2025

[38] [38]

Ryan Greenblatt, Carson Denison, Benjamin Wright, Fabien Roger, Monte MacDiarmid, Sam Marks, Johannes Treutlein, Tim Belonax, Jack Chen, David Duvenaud, Akbir Khan, Julian Michael, Sören Mindermann, Ethan Perez, Linda Petrini, Jonathan Uesato, Jared Kaplan, Buck Shlegeris, Samuel R. Bow- man, and Evan Hubinger. 2024. Alignment faking in large language mod...

View on Pith Pith review arXiv 2024

[39] [39]

Luke Guerdan, Solon Barocas, Kenneth Holstein, Hanna Wallach, Zhiwei Steven Wu, and Alexandra Chouldechova. 2025. Validating LLM-as-a-Judge Systems under Rating Indeterminacy. InAdvances in Neural Information Processing Systems, Vol. 38. https://proceedings.neurips.cc/paper_files/paper/2025/file/ a309239c11a28c597d050bd4a1752d32-Paper-Conference.pdf

View on Pith 2025

[40] [40]

Neel Guha, Julian Nyarko, Daniel E. Ho, Christopher Ré, Adam Chilton, Aditya Narayana, Alex Chohlas-Wood, Austin Peters, Brandon Waldon, Daniel N. Rockmore, Diego Zambrano, Dmitry Talisman, Enam Hoque, Faiz Surani, Frank Fagan, Galit Sarfaty, Gregory M. Dickinson, Haggai Porat, Jason Hegland, Jes- sica Wu, Joe Nudell, Joel Niklaus, John Nay, Jonathan H. C...

View on Pith

[41] [41]

Michael V Heinz, Daniel M Mackin, Brianna M Trudeau, Sukanya Bhattacharya, Yinzhou Wang, Haley A Banta, Abi D Jewett, Abigail J Salzhauer, Tess Z Griffin, and Nicholas C Jacobson. 2025. Randomized trial of a generative ai chatbot for mental health treatment.NEJM AI2, 4 (2025), AIoa2400802

View on Pith 2025

[42] [42]

Natali Helberger, Jo Pierson, and Thomas Poell. 2018. Governing Online Plat- forms: From Contested to Cooperative Responsibility.The Information Society 34, 1 (Jan. 2018), 1–14. https://doi.org/10.1080/01972243.2017.1391913

View on Pith doi:10.1080/01972243.2017.1391913 2018

[43] [43]

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021. Measuring Massive Multitask Language Understanding. InInternational Conference on Learning Representations. https: //openreview.net/forum?id=d7KBjmI3GmQ

View on Pith 2021

[44] [44]

Kashmir Hill and Jennifer Valentino-DeVries. 2025. What OpenAI Did When ChatGPT users lost touch with reality.The New York Times. https://www. nytimes. com/2025/11/23/technology/openai-chatgpt-users-risks. html(2025)

View on Pith 2025

[45] [45]

Elizabeth Hilliard, Akshaya Jagadeesh, Alex Cook, Steele Billings, Nicholas Skyt- land, Alicia Llewellyn, Jackson Paull, Nathan Paull, Nolan Kurylo, Keatra Nesbitt, Robert Gruenewald, Anthony Jantzi, and Omar Chavez. 2025. Measuring AI Alignment with Human Flourishing. https://doi.org/10.48550/arXiv.2507.07787 arXiv:2507.07787 [cs.AI]

View on Pith Pith review arXiv doi:10.48550/arxiv.2507.07787 2025

[46] [46]

Sarah Holland, Ahmed Hosny, Sarah Newman, Joshua Joseph, and Kasia Chmielinski. 2020. The dataset nutrition label.Data protection and privacy12, 12 (2020), 1

View on Pith 2020

[47] [47]

Jessica Hullman and Nicholas Diakopoulos. 2011. Visualization Rhetoric: Fram- ing Effects in Narrative Visualization.IEEE Transactions on Visualization and Computer Graphics17, 12 (2011), 2231–2240. https://doi.org/10.1109/TVCG. 2011.255

View on Pith doi:10.1109/tvcg 2011

[48] [48]

Lujain Ibrahim, Canfer Akbulut, Rasmi Elasmar, Charvi Rastogi, Minsuk Kahng, Meredith Morris, Kevin McKee, Verena Rieser, Murray Shanahan, and Laura Weidinger. 2026. Multi-turn Evaluation of Anthropomorphic Be- haviours in Large Language Models. InInternational Conference on Learn- ing Representations. https://proceedings.iclr.cc/paper_files/paper/2026/ha...

View on Pith 2026

[49] [49]

Lujain Ibrahim, Saffron Huang, Lama Ahmad, Umang Bhatt, and Markus An- derljung. 2025. Towards Interactive Evaluations for Interaction Harms in Human-AI Systems. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, Vol. 8. 1302–1310. https://doi.org/10.1609/aies.v8i2.36631

View on Pith doi:10.1609/aies.v8i2.36631 2025

[50] [50]

Institute for Integrated Transitions and Collective Intelligence Project

View on Pith

[51] [51]

Abigail Z. Jacobs and Hanna Wallach. 2021. Measurement and Fairness. InProceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. Association for Computing Machinery. https://doi.org/10.1145/ 3442188.3445901

View on Pith 2021

[52] [52]

Kiana Jafari Meimandi, Paul Ulrich Nikolaus Rust, Duncan Eddy, Robbie Fraser, Nina Vasan, Darja Djordjevic, Akanksha Dadlani, Max Lamparth, Eugenia Kim, and Mykel Kochenderfer. 2026. Expert Evaluation and the Limits of Human Feedback in Mental Health AI Safety Testing. InThe 2026 ACM Conference on Fairness, Accountability, and Transparency. 3894–3920

View on Pith 2026

[53] [53]

Lucie-Aimée Kaffee, Giada Pistilli, and Yacine Jernite. 2026. INTIMA: A Bench- mark for Human-AI Companionship Behavior. InInternational Conference on Learning Representations. https://proceedings.iclr.cc/paper_files/paper/2026/ hash/05b12f103c9e613efc4c85674cdc9066-Abstract-Conference.html Open Benchmark of AI Impact on Humans (ImpactBench)

View on Pith 2026

[54] [54]

Celeste Kidd and Abeba Birhane. 2023. How AI can distort human beliefs. Science380, 6651 (2023), 1222–1223

View on Pith 2023

[55] [55]

David E. Kieras and Susan Bovair. 1984. The Role of a Mental Model in Learning to Operate a Device.Cognitive Science8, 3 (1984), 255–273. https://doi.org/10. 1207/s15516709cog0803_3

View on Pith 1984

[56] [56]

Pilyoung Kim, Yun Xie, and Sujin Yang. 2026. “I am here for you”: Adolescents’ perceptions of relational conversational AI and the social-emotional correlates of preference.Computers in Human Behavior Reports(2026), 101297. https: //doi.org/10.1016/j.chbr.2026.101297

View on Pith doi:10.1016/j.chbr.2026.101297 2026

[57] [57]

Tae Soo Kim, Yoonjoo Lee, Jamin Shin, Young-Ho Kim, and Juho Kim. 2024. EvalLM: Interactive Evaluation of Large Language Model Prompts on User- Defined Criteria. InProceedings of the CHI Conference on Human Factors in Computing Systems. Association for Computing Machinery, Article 306, 21 pages. https://doi.org/10.1145/3613904.3642216

View on Pith doi:10.1145/3613904.3642216 2024

[58] [58]

Artur Klingbeil, Cassandra Grützner, and Philipp Schreck. 2024. Trust and reliance on AI—An experimental study on the extent and costs of overreliance on AI.Computers in Human Behavior160 (2024), 108352. https://doi.org/10. 1016/j.chb.2024.108352

View on Pith 2024

[59] [59]

KORA Benchmark. 2026. KORA Benchmark: The Benchmark for AI Child Safety. https://korabench.ai. Accessed: 2026-04-10

View on Pith 2026

[60] [60]

Nataliya Kosmyna, Eugene Hauptmann, Ye Tong Yuan, Jessica Situ, Xian-Hao Liao, Ashly Vivian Beresnitzky, Iris Braunstein, and Pattie Maes. 2025. Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task.arXiv preprint arXiv:2506.088724 (2025), 206

View on Pith arXiv 2025

[61] [61]

David Kreitmeir and Paul A Raschky. 2024. The Heterogeneous Productivity Effects of Generative AI.arXiv preprint arXiv:2403.01964(2024)

View on Pith Pith review arXiv 2024

[62] [62]

Nomisha Kurian. 2023. AI’s Empathy Gap: The Risks of Conversational Artificial Intelligence for Young Children’s Well-Being and Key Ethical Considerations for Early Childhood Education and Care.Contemporary Issues in Early Childhood26, 1 (2023), 132–139. https://doi.org/10.1177/14639491231206004 First published online in 2023

View on Pith doi:10.1177/14639491231206004 2023

[63] [63]

Philippe Laban, Hiroaki Hayashi, Yingbo Zhou, and Jennifer Neville. 2025. Llms get lost in multi-turn conversation.arXiv preprint arXiv:2505.06120(2025)

View on Pith Pith review arXiv 2025

[64] [64]

Ruben Laukkonen, Seb Krier, Chloé Bakalar, Shamil Chandaria, Morten Kringel- bach, Adam Elwood, Daniel Ford, Fernando Rosas, Maty Bohacek, Matija Franklin, et al . 2026. Positive alignment: Artificial intelligence for human flourishing.arXiv preprint arXiv:2605.10310(2026)

View on Pith Pith review arXiv 2026

[65] [65]

Hao-Ping Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. 2025. The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers. InProceedings of the 2025 CHI conference on human factors in computing systems. 1–22

View on Pith 2025

[66] [66]

Hao-Ping Lee, Advait Sarkar, Lev Tankelevitch, Ian Drosos, Sean Rintel, Richard Banks, and Nicholas Wilson. 2025. The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers. InProceedings of the 2025 CHI Conference on Human Factors in Computing Systems. Associati...

View on Pith doi:10.1145/3706598 2025

[67] [67]

Joanne Leong, Pat Pataranutaporn, Valdemar Danry, Florian Perteneder, Yaoli Mao, and Pattie Maes. 2024. Putting Things into Context: Generative AI-Enabled Context Personalization for Vocabulary Learning Improves Learning Motivation. InProceedings of the CHI Conference on Human Factors in Computing Systems. 1–15

View on Pith 2024

[68] [68]

Jennifer S Lerner and Philip E Tetlock. 1999. Accounting for the effects of accountability.Psychological bulletin125, 2 (1999), 255

View on Pith 1999

[69] [69]

Robert O Lewis. 1992.Independent verification and validation: A life cycle engineering process for quality software. John Wiley & Sons

View on Pith 1992

[70] [70]

Yahan Li, Jifan Yao, John Bosco S. Bunyi, Adam C. Frank, Angel Hsing-Chi Hwang, and Ruishan Liu. 2025. CounselBench: A Large-Scale Expert Eval- uation and Adversarial Benchmarking of Large Language Models in Men- tal Health Question Answering. https://doi.org/10.48550/arXiv.2506.08584 arXiv:2506.08584 Version 4, revised May 2026

View on Pith Pith review arXiv doi:10.48550/arxiv.2506.08584 2025

[71] [71]

Bill Yuchen Lin, Yuntian Deng, Khyathi Chandu, Faeze Brahman, Abhilasha Ravichander, Valentina Pyatkin, Nouha Dziri, Ronan Le Bras, and Yejin Choi

View on Pith

[72] [72]

Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. TruthfulQA: Measuring How Models Mimic Human Falsehoods. InProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, Dublin, Ireland, 3214–3252. https: //doi.org/10.18653/v1/2022.acl-long.229

View on Pith doi:10.18653/v1/2022.acl-long.229 2022

[73] [73]

Auren R. Liu, Pat Pataranutaporn, and Pattie Maes. 2025. The Heteroge- neous Effects of AI Companionship: An Empirical Model of Chatbot Usage and Loneliness and a Typology of User Archetypes. InProceedings of the AAAI/ACM Conference on AI, Ethics, and Society, Vol. 8. 1585–1597. https: //doi.org/10.1609/aies.v8i2.36658

View on Pith doi:10.1609/aies.v8i2.36658 2025

[74] [74]

Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2511–2522. https://doi. org/10.18653/v1/2023.emnlp-main.153

View on Pith doi:10.18653/v1/2023.emnlp-main.153 2023

[75] [75]

Kim Malfacini. 2025. The Impacts of Companion AI on Human Relationships: Risks, Benefits, and Design Considerations.AI & Society40 (2025), 5527–5540. https://doi.org/10.1007/s00146-025-02318-6

View on Pith doi:10.1007/s00146-025-02318-6 2025

[76] [76]

Mantas Mazeika, Long Phan, Xuwang Yin, Andy Zou, Zifan Wang, Norman Mu, Elham Sakhaee, Nathaniel Li, Steven Basart, Bo Li, David Forsyth, and Dan Hendrycks. 2024. HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal. https://doi.org/10.48550/arXiv. 2402.04249 arXiv:2402.04249

View on Pith Pith review arXiv doi:10.48550/arxiv.2402.04249 2024

[77] [77]

Niloofar Mikaeili, Mahdi Naeim, and Mohammad Narimani. 2025. Reimagining mental health with artificial intelligence: early detection, personalized care, and a preventive ecosystem.Journal of Multidisciplinary Healthcare(2025), 7355–7373

View on Pith 2025

[78] [78]

Robert J. Mislevy, Russell G. Almond, and Janice F. Lukas. 2004.A Brief Introduc- tion to Evidence-Centered Design. CSE Report 632. National Center for Research on Evaluation, Standards, and Student Testing, University of California, Los Angeles. https://cresst.org/wp-content/uploads/R632.pdf

View on Pith 2004

[79] [79]

Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasser- man, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, and Timnit Gebru

View on Pith

[80] [80]

Jared Moore, Declan Grabb, William Agnew, Kevin Klyman, Stevie Chancellor, Desmond C Ong, and Nick Haber. 2025. Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers.. In Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Trans- parency. 599–627

View on Pith 2025

[81] [81]

Ehsan Namaziandost and Mark Feng Teng. 2026. AI-Supported Socratic Dia- logue Agents vs. Generic AI Chatbots for Intermediate EFL Learners’ Critical Thinking, Epistemic Beliefs, and Argumentation: Mixed-Methods Evidence. International Journal of Applied Linguistics(2026). https://doi.org/10.1111/ijal. 70268

View on Pith doi:10.1111/ijal 2026

[82] [82]

Chiara Natali, Luca Marconi, Leslye Denisse Dias Duran, and Federico Cabitza

View on Pith

[83] [83]

Joe Needham, Giles Edkins, Govind Pimpale, Henning Bartsch, and Marius Hobbhahn. 2025. Large Language Models Often Know When They Are Being Evaluated. arXiv:2505.23836 [cs.CL] https://arxiv.org/abs/2505.23836

View on Pith Pith review arXiv 2025

[84] [84]

Jakob Nielsen. 1994. 10 Usability Heuristics for User Interface Design. Nielsen Norman Group. https://www.nngroup.com/articles/ten-usability-heuristics/ Last reviewed January 30, 2024; accessed September 22, 2026

View on Pith 1994

[85] [85]

Bayode Ogunleye, Kudirat Ibilola Zakariyyah, Oluwaseun Ajao, Olakunle Olayinka, and Hemlata Sharma. 2024. A Systematic Review of Generative AI for Teaching and Learning Practice.Education Sciences14, 6 (2024), 636. https://doi.org/10.3390/educsci14060636

View on Pith doi:10.3390/educsci14060636 2024

[86] [86]

Juhyun Oh, Inha Cha, Michael Saxon, Hyunseung Lim, Shaily Bhatt, and Al- ice Oh. 2025. Culture is Everywhere: A Call for Intentionally Cultural Eval- uation. InFindings of the Association for Computational Linguistics: EMNLP

View on Pith 2025

[87] [87]

AI-induced deskilling in medicine: a mixed-method review and research agenda for healthcare and beyond.Artificial Intelligence Review58, 11 (2025), 356

View on Pith 2025

[88] [88]

Pat Pataranutaporn. 2024.Cyborg Psychology: The Art & Science of Design- ing Human-AI Systems that Support Human Flourishing. Ph. D. Dissertation. Massachusetts Institute of Technology

View on Pith 2024

[89] [89]

Pat Pataranutaporn, Joanne Leong, Valdemar Danry, Alyssa P. Lawson, Pattie Maes, and Misha Sra. 2022. AI-Generated Virtual Instructors Based on Liked or Admired People Can Improve Motivation and Foster Positive Emotions for Learning. InIEEE Frontiers in Education Conference (FIE). IEEE, 1–9. https: //doi.org/10.1109/FIE56618.2022.9962478

View on Pith doi:10.1109/fie56618.2022.9962478 2022

[90] [90]

Pat Pataranutaporn, Kavin Winson, Peggy Yin, Auttasak Lapapirojn, Pichayoot Ouppaphan, Monchai Lertsutthiwong, Pattie Maes, and Hal Hershfield. 2024. Future You: A Conversation with an AI-Generated Future Self Reduces Anxi- ety, Negative Emotions, and Increases Future Self-Continuity.arXiv preprint arXiv:2405.12514(2024)

View on Pith Pith review arXiv 2024

[91] [91]

Ethan Perez, Saffron Huang, Francis Song, Trevor Cai, Roman Ring, John Aslanides, Amelia Glaese, Nat McAleese, and Geoffrey Irving. 2022. Red Team- ing Language Models with Language Models. InProceedings of the 2022 Con- ference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics. https://doi.org/10.18653/v1/2022...

View on Pith doi:10.18653/v1/2022.emnlp-main.225 2022

[92] [92]

Association for Computational Linguistics, Suzhou, China, 19156–19168. https://doi.org/10.18653/v1/2025.findings-emnlp.1043

View on Pith doi:10.18653/v1/2025.findings-emnlp.1043 2025

[93] [93]

OpenAI. 2026.GPT-6 Sol and GPT-6 Luna: October 2026 Update. System Card. OpenAI. https://deploymentsafety.openai.com/gpt-6-october

View on Pith 2026

[94] [94]

Inioluwa Deborah Raji, Emily Denton, Emily M. Bender, Alex Hanna, and Amandalynne Paullada. 2021. AI and the Everything in the Whole Wide World Benchmark. InProceedings of the Neural Infor- mation Processing Systems Track on Datasets and Benchmarks, Vol. 1. https://datasets-benchmarks-proceedings.neurips.cc/paper/2021/hash/ 084b6fbb10729ed4da8c3d3f5a3ae7c...

View on Pith 2021

[95] [95]

Inioluwa Deborah Raji, Peggy Xu, Colleen Honigsberg, and Daniel E. Ho

View on Pith

[96] [96]

Anku Rani, Valdemar Danry, Paul Pu Liang, Andrew Lippman, and Pattie Maes

View on Pith

[97] [97]

David Rein, Betty Li Hou, Asa Cooper Stickland, Jackson Petty, Richard Yuanzhe Pang, Julien Dirani, Julian Michael, and Samuel R Bowman. 2024. Gpqa: A graduate-level google-proof q&a benchmark. InFirst Conference on Language Modeling

View on Pith 2024

[98] [98]

Stephen R. Pfohl, Heather Cole-Lewis, Rory Sayres, et al. 2024. A toolbox for surfacing health equity harms and biases in large language models.Nature Medicine30 (2024), 3590–3600. https://doi.org/10.1038/s41591-024-03258-2 Open Benchmark of AI Impact on Humans Team

View on Pith doi:10.1038/s41591-024-03258-2 2024

[99] [99]

Rachel Poonsiriwong, Chayapatr Archiwaranguprok, Constanze Albrecht, Peggy Yin, Nattavudh Powdthavee, Hal Hershfield, Monchai Lertsutthiwong, Kavin Winson, and Pat Pataranutaporn. 2026. Simulating Life Paths with Digital Twins: AI-Generated Future Selves Influence Decision-Making and Expand Hu- man Choice. InProceedings of the Augmented Humans Internation...

View on Pith 2026

[100] [100]

Andrew Rule, Cody Abbey, Huan Wang, Scott Rozelle, and Manpreet K. Singh

View on Pith

[101] [101]

Richard M Ryan and Edward L Deci. 2000. Self-determination theory and the facilitation of intrinsic motivation, social development, and well-being. American psychologist55, 1 (2000), 68

View on Pith 2000

[102] [102]

Richard M Ryan and Edward L Deci. 2001. On happiness and human potentials: A review of research on hedonic and eudaimonic well-being.Annual review of psychology52, 1 (2001), 141–166

View on Pith 2001

[103] [103]

Julio Saez-Rodriguez, Philipp Sven Lars Schäfer, Nikolas Kalavros, and Gustavo Stolovitzky. 2026. Benchmarking biomedical foundation models.Nature Methods 23, 9 (2026), 1724–1733. https://doi.org/10.1038/s41592-026-03182-y

View on Pith doi:10.1038/s41592-026-03182-y 2026

[104] [104]

Jonne Sälevä, Duygu Ataman, and Constantine Lignos. 2025. Beyond statistical significance: Quantifying uncertainty and statistical variability in multilingual and multitask NLP evaluation. InProceedings of the 14th International Joint Con- ference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Comp...

View on Pith 2025

[105] [105]

Shalom H Schwartz. 2012. An overview of the Schwartz theory of basic values. Online readings in Psychology and Culture2, 1 (2012)

View on Pith 2012

[106] [106]

Anka Reuel, Amelia Hardy, Chandler Smith, Max Lamparth, Malcolm Hardy, and Mykel J. Kochenderfer. 2024. BetterBench: Assessing AI Benchmarks, Uncover- ing Issues, and Establishing Best Practices. InAdvances in Neural Information Processing Systems, Vol. 37. https://doi.org/10.52202/079017-0685

View on Pith doi:10.52202/079017-0685 2024

[107] [107]

Marco Tulio Ribeiro, Tongshuang Wu, Carlos Guestrin, and Sameer Singh

View on Pith

[108] [108]

Gregory Serapio-García, Mustafa Safdari, Clément Crepy, Luning Sun, Stephen Fitz, Peter Romero, Marwa Abdulhai, Aleksandra Faust, and Maja Matarić. 2025. A psychometric framework for evaluating and shaping personality traits in large language models.Nature Machine Intelligence7, 12 (2025), 1954–1968. https://doi.org/10.1038/s42256-025-01115-6

View on Pith doi:10.1038/s42256-025-01115-6 2025

[109] [109]

Kathrin Seßler, Ozan Kepir, and Enkelejda Kasneci. 2024. Enhancing student motivation through LLM-powered learning environments: a comparative study. InEuropean Conference on Technology Enhanced Learning. Springer, 156–162

View on Pith 2024

[110] [110]

Measurement of flourishing: a scoping review.Frontiers in Psychology15 (2024), 1293943. https://doi.org/10.3389/fpsyg.2024.1293943

View on Pith doi:10.3389/fpsyg.2024.1293943 2024

[111] [111]

Ben Shneiderman. 1996. The Eyes Have It: A Task by Data Type Taxonomy for Information Visualizations. InProceedings of the 1996 IEEE Symposium on Visual Languages. IEEE, 336–343. https://doi.org/10.1109/VL.1996.545307

View on Pith doi:10.1109/vl.1996.545307 1996

[112] [112]

Molly G. Smith, Thomas N. Bradbury, and Benjamin R. Karney. 2025. Can Generative AI Chatbots Emulate Human Connection? A Relationship Science Perspective.Perspectives on Psychological Science20, 6 (2025), 1081–1099. https: //doi.org/10.1177/17456916251351306

View on Pith doi:10.1177/17456916251351306 2025

[113] [113]

Nicolas Spatola. 2024. The efficiency-accountability tradeoff in AI integration: Effects on human performance and over-reliance.Computers in Human Behavior: Artificial Humans2, 2 (2024), 100099

View on Pith 2024

[114] [114]

Rakshith S Srinivasa, Zora Che, Chen Bo Calvin Zhang, Diego Mares, Ernesto Hernandez, Jayeon Park, Dean Lee, Guillermo Mangialardi, Charmaine Ng, Ed-Yeremai Hernandez Cardona, Anisha Gunjal, Yunzhong He, Bing Liu, and Chen Xing. 2025. TutorBench: A Benchmark to Assess Tutoring Capabili- ties of Large Language Models. https://doi.org/10.48550/arXiv.2510.02...

View on Pith doi:10.48550/arxiv.2510.02663 2025

[115] [115]

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. 2023. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models.Transactions on machine learning research(2023)

View on Pith 2023

[116] [116]

Ivan Sekulic, Silvia Terragni, Victor Guimarães, Nghia Khau, Bruna Guedes, Modestas Filipavicius, Andre Ferreira Manso, and Roland Mathis. 2024. Reliable LLM-based User Simulator for Task-Oriented Dialogue Systems. InProceedings of the 1st Workshop on Simulating Conversational Intelligence in Chat (SCI-CHAT 2024). Association for Computational Linguistics...

View on Pith 2024

[117] [117]

Andrew D Selbst, Danah Boyd, Sorelle A Friedler, Suresh Venkatasubramanian, and Janet Vertesi. 2019. Fairness and abstraction in sociotechnical systems. InProceedings of the conference on fairness, accountability, and transparency. 59–68

View on Pith 2019

[118] [118]

Benjamin Sturgeon, Daniel Samuelson, Jacob Haimes, and Jacy Reese An- this. 2025. HumanAgencyBench: Scalable Evaluation of Human Agency Support in AI Assistants. https://doi.org/10.48550/arXiv.2509.08494 arXiv:2509.08494 [cs.CY]

View on Pith Pith review arXiv doi:10.48550/arxiv.2509.08494 2025

[119] [119]

Xiaoran Sun, Yunqi Wang, and Brandon T. McDaniel. 2026. AI companions and adolescent social relationships: Benefits, risks, and bidirectional influences. Child Development Perspectives20, 2 (2026), 91–98. https://doi.org/10.1093/ cdpers/aadaf009

View on Pith 2026

[120] [120]

Shruthi Shekar, Pat Pataranutaporn, Chethan Sarabu, Guillermo A Cecchi, and Pattie Maes. 2025. People Overtrust AI-Generated Medical Advice despite Low Accuracy.NEJM AI(2025), AIoa2300015

View on Pith 2025

[121] [121]

Tyler J VanderWeele. 2017. On the promotion of human flourishing.Proceedings of the National Academy of Sciences114, 31 (2017), 8148–8156

View on Pith 2017

[122] [122]

Tyler J VanderWeele, Byron R Johnson, Piotr T Bialowolski, Rebecca Bonhag, Matt Bradshaw, Thomas Breedlove, Brendan Case, Ying Chen, Zhuo Job Chen, Victor Counted, et al. 2025. The Global Flourishing Study: Study profile and initial results on flourishing.Nature Mental Health(2025), 1–18

View on Pith 2025

[123] [123]

Tyler J VanderWeele and Jonathan D Teubner. 2026. Flourishing Considerations for AI.Information17, 1 (2026), 88

View on Pith 2026

[124] [124]

Iris Vessey. 1991. Cognitive Fit: A Theory-Based Analysis of the Graphs Versus Tables Literature.Decision Sciences22, 2 (1991), 219–240. https://doi.org/10. 1111/j.1540-5915.1991.tb00344.x

View on Pith 1991

[125] [125]

Lev S. Vygotsky. 1978.Mind in Society: The Development of Higher Psychological Processes. Harvard University Press, Cambridge, MA

View on Pith 1978

[126] [126]

Stanford Center for Digital Health and Stanford Social Media Lab. 2026. AI Safety: Defining & Measuring Potential Harms of Chatbots. Workshop Report. Stanford University. https://cdh.stanford.edu/sites/g/files/sbiybj29486/ files/media/file/ai-safety-defining-and-measuring-potential-harms-of- chatbots_final.pdf

View on Pith 2026

[127] [127]

Stanford HAI and Collective Intelligence Project. 2025. Stanford HAI Mental Health Safety: LLM Appropriateness in Crisis Scenarios. Weval evaluation blueprint. https://weval.org/analysis/stanford-hai-mental-health-safety-eval/ 7a8bca546abd3b83/2025-10-05T10-40-31-897Z

View on Pith 2025

[128] [128]

Jin Wang and Wenxiang Fan. 2025. The effect of ChatGPT on students’ learning performance, learning perception, and higher-order thinking: insights from a meta-analysis.Humanities and Social Sciences Communications12, 1 (2025), 1–21

View on Pith 2025

[129] [129]

Yuqing Wang and Yun Zhao. 2024. Rupbench: Benchmarking reasoning under perturbations for robustness evaluation in large language models.arXiv preprint arXiv:2406.11020(2024)

View on Pith Pith review arXiv 2024

[130] [130]

John Torous and Eric J Topol. 2025. Assessing generative artificial intelligence for mental health.The Lancet406, 10504 (2025), 683

View on Pith 2025

[131] [131]

Laura Weidinger, Inioluwa Deborah Raji, Hanna Wallach, Margaret Mitchell, Angelina Wang, Olawale Salaudeen, Rishi Bommasani, Sayash Kapoor, Deep Ganguli, Sanmi Koyejo, and William Isaac. 2025. Toward an Evaluation Science for Generative AI Systems.arXiv preprint arXiv:2503.05336(2025). https: //doi.org/10.48550/arXiv.2503.05336

View on Pith Pith review arXiv doi:10.48550/arxiv.2503.05336 2025

[132] [132]

Laura Weidinger, Maribeth Rauh, Nahema Marchal, Arianna Manzini, Lisa Anne Hendricks, Juan Mateos-Garcia, Stevie Bergman, Jackie Kay, Conor Griffin, Ben Bariach, et al. 2023. Sociotechnical safety evaluation of generative ai systems. arXiv preprint arXiv:2310.11986(2023). Open Benchmark of AI Impact on Humans (ImpactBench)

View on Pith Pith review arXiv 2023

[133] [133]

David Wood, Jerome S. Bruner, and Gail Ross. 1976. The Role of Tutoring in Problem Solving.Journal of Child Psychology and Psychiatry17, 2 (1976), 89–100. https://doi.org/10.1111/j.1469-7610.1976.tb00381.x

View on Pith doi:10.1111/j.1469-7610.1976.tb00381.x 1976

[134] [134]

Harry Yizhou Tian, Hasan Amin, and Ming Yin. 2026. Understanding the effects of AI-assisted critical thinking on human-AI decision making. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–30

View on Pith 2026

[135] [135]

Chunpeng Zhai, Santoso Wibowo, and Lily D Li. 2024. The effects of over- reliance on AI dialogue systems on students’ cognitive abilities: a systematic review.Smart Learn. Environ.11, 1 (June 2024)

View on Pith 2024

[136] [136]

Guanhua Wang, Wenna Wang, Daozhou Yang, and Jifan Ren. 2026. Generative AI, Cognitive Offloading, and Learner Agency in Higher Education: A Scop- ing Review.Behavioral Sciences16, 7 (2026), 1150. https://doi.org/10.3390/ bs16071150

View on Pith 2026

[137] [137]

Jennifer Wang, Joachim Baumann, Daniel E. Ho, and Sanmi Koyejo. 2026. API Benchmark Scores Do Not Reliably Transfer to Chatbot Interfaces. arXiv:2609.08861 [cs.AI] https://arxiv.org/abs/2609.08861

View on Pith Pith review arXiv 2026

[138] [138]

Yutong Zhang, Dora Zhao, Yixin Wang, Rebecca Anselmetti, Jeffrey T Hancock, Robert Kraut, and Diyi Yang. 2026. Living with AI Companions: Sustained AI Companionship Predicts Lower Well-Being Through Lower Human Interaction. arXiv preprint arXiv:2609.07243(2026)

View on Pith Pith review arXiv 2026

[139] [139]

Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM-as-a-Judge with MT- Bench and Chatbot Arena. InAdvances in Neural Information Processing Systems, Vol. 36. https://arxiv.org/abs/2306.05685

View on Pith Pith review arXiv 2023

[140] [140]

Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed Chi, Quoc V. Le, and Denny Zhou. 2022. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models. InAdvances in Neural Information Processing Systems, Vol. 35. https://proceedings.neurips.cc/paper/2022/hash/ 9d5609613524ecf4f15af0f7b31abca4-Abstract-Conference.html

View on Pith 2022

[141] [146]

Yuhao Zhang, Eng Lieh Ouh, Adam Ho, Siaw Ling Lo, Kar Way Tan, and Feng Lin. 2025. PromptTutor: Effects of an LLM-based chatbot on learning outcomes and motivation in flipped classrooms. InProceedings of the 30th ACM Conference on Innovation and Technology in Computer Science Education V. 1. 445–451

View on Pith 2025

[142] [147]

Yutong Zhang, Dora Zhao, Jeffrey T Hancock, Robert Kraut, and Diyi Yang

View on Pith

[143] [148]

Interaction with AI companions and psychological well-being.Nature human behaviour(2026), 1–15

View on Pith 2026

[144] [151]

Qiuhan Zhu, Xiangnan Li, Yiang Dong, Pengcheng Chang, and Mengmeng Fan. 2026. Not all cognitive offloading is equal: distinguishing dependent and autonomous offloading to generative AI.Frontiers in Psychology17 (2026), 1878629. https://doi.org/10.3389/fpsyg.2026.1878629 A Additional Platform Views The main text presents the platform components directly re...

View on Pith doi:10.3389/fpsyg.2026.1878629 2026

[145] [2019]

Model cards for model reporting. InProceedings of the conference on fairness, accountability, and transparency. 220–229

View on Pith

[146] [2020]

Beyond Accuracy: Behavioral Testing of NLP Models with CheckList. InProceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 4902–4912. https://doi. org/10.18653/v1/2020.acl-main.442

View on Pith doi:10.18653/v1/2020.acl-main.442

[147] [2022]

Outsider Oversight: Designing a Third Party Audit Ecosystem for AI Governance. (2022). https://doi.org/10.48550/arXiv.2206.04737 arXiv:2206.04737 [cs.CY]

View on Pith Pith review arXiv doi:10.48550/arxiv.2206.04737 2022

[148] [2023]

LegalBench: A Collaboratively Built Benchmark for Measuring Legal Rea- soning in Large Language Models. https://doi.org/10.48550/arXiv.2308.11462 arXiv:2308.11462 [cs.CL]

View on Pith Pith review arXiv doi:10.48550/arxiv.2308.11462

[149] [2024]

WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild. https://doi.org/10.48550/arXiv.2406.04770 arXiv:2406.04770

View on Pith Pith review arXiv doi:10.48550/arxiv.2406.04770

[150] [2025]

IFIT Conflict Resolution Scenarios (2025). Weval evaluation blue- print. https://weval.org/analysis/ifit-conflict-resolution/3bdd4deec7f6cc04/ 2025-10-05T12-54-16-403Z Run of 5 October 2025

View on Pith 2025

[151] [2026]

Dialogues with AI Reduce Beliefs in Misinformation but Build No Lasting Discernment Skills. InProceedings of the 2026 CHI Conference on Human Factors in Computing Systems. 1–26

View on Pith 2026