Pith. sign in

REVIEW 3 major objections 4 minor 81 references

LLM-generated personalized nudges for improving pro-environmental behavior: Field evidence from resource conservation

T0 review · 3 major / 4 minor · reviewed 2026-08-02 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that LLM-personalized nudges, iteratively tailored to each household's usage and interaction history, reduce electricity consumption by 0.56 kWh per room-day versus text-only feedback—an 18.3 percentage-point gain in savin

desk verdict A genuine field RCT on LLM-personalized nudges, but the headline electricity effect rests on self-entered meter readings; referee it with a demand for independent measurement. read the letter →

arxiv 2604.03881 v2 pith:5UKUGOYU submitted 2026-04-04 cs.CY cs.AIcs.HC

Zonghan Li , Yi Liu , Chunyan Wang , Song Tong , Kaiping Peng , Feng Ji This is my paper · ORCID
classification cs.CYcs.AIcs.HC
keywords largelanguagemodelspersonalizednudgesfieldexperimentelectricityconservationhotwaterrandomizedcontrolledtrialbehaviorchangeiterativepersonalization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether large language models can take over the cognitive work that conventional feedback nudges leave to the individual: translating 'you used more than your peer group' into 'here is one thing you can do tomorrow.' In a randomized field trial with 233 dormitory residents, it compares standard text feedback, image-enhanced feedback, and feedback plus weekly LLM-generated personalized suggestions, scenarios, and savings estimates. It finds that the LLM-personalized arm reduced electricity consumption by 0.56 kWh per room-day relative to text feedback (p=0.014), an 18.3 percentage-point higher adjusted saving rate, with the gain appearing in the first two rounds and persisting. Image-enhanced formatting alone showed no such benefit. The paper argues this is field evidence that iterative LLM personalization can move objectively measured behavior, while also showing that behavioral friction (hot water versus electricity) may set a boundary on the effect.

What carries the argument

Iterative personalization. Each week, a retrieval-augmented LLM agent receives the participant's demographic and psychological profile, consumption data through the previous round, prior nudge suggestions, and any explicit feedback; it updates the profile in a chain-of-thought reasoning step, then generates a nudge that pairs the standard usage report with (1) personalized suggestions, (2) a behavioral scenario embedded in the participant's routines, and (3) a quantitative analogy of expected savings. The key work of this machinery is to convert raw usage statistics into a concrete, situation-specific next action, and to revise that action as the participant's circumstances and behavior evol

What would settle it

Compare each participant's self-entered daily electricity and hot-water values against the independent meter and billing records held by the facility for all five intervention weeks. If the discrepancy rate is higher in the LLM-personalized arm and the 0.56 kWh contrast disappears once reconciled values are used, the causal claim fails; alternatively, a pre-registered replication with automatic meter reading would settle it in one field season.

Watch

Extended reading notes

Core claim

The central discovery is an empirical contrast: personalized nudges produced by an LLM that updates each participant's profile weekly outperform static text or image feedback on a measured behavior. In covariate-adjusted models, predicted electricity use was 2.03 kWh per room-day in the LLM-personalized group versus 2.58 in the text group and 2.53 in the image group, corresponding to adjusted saving rates of 32.4%, 14.1%, and 16.0%. The difference between the LLM group and the text group, 0.56 kWh per room-day (p=0.014), is the paper's headline claim; hot-water savings were directionally consistent (9.8 percentage points, p=0.087) but imprecise. The electricity effect emerged in the first tw

Load-bearing premise

The entire treatment effect rests on participants typing their own meter and billing readings into the chatbot; if people in the personalized group submitted lower numbers because the messages reshaped what they thought was expected, the measured savings would be inflated even if actual behavior did not change.

Editorial extensions

If this is right

  • If the electricity effect replicates, utilities and campus managers could add LLM personalization to existing meter-feedback programs and achieve conservation gains of roughly 18 percentage points in saving rate without changing prices or incentives.
  • The early, persistent gain for electricity while hot-water savings attenuated implies the design is most promising for low-friction behaviors; for comfort-linked behaviors, gains may fade or require additional support.
  • Because the personalized arm also had higher engagement and more task-focused use, message actionability rather than visual appeal is likely the active ingredient; image enhancement alone did not work.
  • The observed mechanism—prospective, planning-oriented content that is updated weekly—aligns with a scalable lightweight-coaching model that could transfer to other repeated decisions such as water use, waste sorting, or energy use in non-residential settings.

Reading between the lines

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

  • We infer that the strongest test would be a replication in which meters and billing records are read automatically rather than self-entered; if the 0.56 kWh effect survives that, the case for deploying LLM-generated nudges at scale is strong.
  • The paper leaves the active ingredient bundled: personalization, scenarios, quantitative analogies, and weekly iteration are confounded; an ablation design—for instance, personalized tips without scenarios, or static rather than updated tips—could identify which piece matters and make the intervention cheaper to run.
  • If selective under-reporting in the personalized arm is ruled out, the results imply that a large share of a nudge's value lies in removing the user's burden of figuring out what to do next—suggesting that even simpler action-oriented prompts might capture part of the effect at lower cost.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper reports a three-arm randomized field experiment (N=233 university dormitory residents in Beijing) testing whether weekly LLM-generated personalized nudges improve electricity and hot-water conservation relative to text-based or image-enhanced conventional feedback. The central claim is that LLM-personalized nudges (T2) reduced electricity consumption by 0.56 kWh per room-day relative to text-based feedback (p=0.014), corresponding to an 18.3 percentage-point higher adjusted saving rate, while hot-water savings were directionally similar but not statistically significant (p=0.087). The paper also provides content analyses, engagement metrics, temporal dynamics, and exploratory heterogeneity analyses. The main result is vulnerable because the outcome data were self-reported by participants through a chatbot and because roughly 27–29% of randomized participants were excluded under researcher-defined rules.

Significance. If the electricity effect is real, this would be a valuable field demonstration that LLM-generated, iteratively personalized nudges can change objectively measured resource-use behavior, with behavioral friction as a plausible boundary condition. The study has notable strengths: a randomized design with cluster-aware assignment, baseline covariates, several robustness checks (permuted-label, cluster-robust, Lee bounds), transparent reporting of exploratory tiers, and detailed content/engagement analyses. However, the central quantitative claim currently rests on self-entered meter readings and shower-cost reports without an independent audit in the main trial. The paper's own Methods acknowledge that the outcome data are self-reported, so the abstract's wording that consumption was 'measured using dormitory meter readings and billing records' overstates the objectivity of the measurement pipeline.

major comments (3)
  1. [Methods, 'Data cleaning'; Results, 'Overall behavioral changes'] The central electricity estimate (0.56 kWh/room-day, p=0.014) is not independently verified. The Methods state: 'Outcome data were self-reported via the chatbot.' For electricity, the internal validation is a redundant weekly aggregation within the same self-report pipeline (<1% flagged), which does not detect consistent under-reporting. For hot water, only ~30% of observations are balance-based, and pilot screenshot verification was not retained in the main trial. Thus the T2-vs-C contrast could be inflated if T2 participants, who were more engaged and received concrete savings targets, systematically under-reported consumption or omitted high-use days. Lee (2009) bounds address selective attrition, not measurement error. I request an independent audit (e.g., facilities-management billing records for the full sample or a large random subset) or a formal sensitivity analysis bounding pla
  2. [Methods, 'Data cleaning'; 'Nudge effect estimation'] The analytic samples exclude a large share of randomized participants: 169 of 233 for electricity and 166 of 233 for hot water. Exclusion criteria include not only >40% missing observations but also 'extreme or unstable consumption patterns,' which are outcome-dependent. If treatment changed behavior in ways that made consumption patterns appear unstable, or if missingness is correlated with treatment engagement, these exclusions can bias the estimated effect. The paper defers arm-specific exclusion counts to Supplementary Tables S1–S3 and reports Lee bounds, but Lee bounds may not cover exclusions based on outcome-dependent instability. Please report the arm-specific exclusion counts in the main text and provide sensitivity analyses that retain all randomized participants (e.g., multiple imputation, inverse-probability weighting, or bounds allowing outcome-dependent selection).
  3. [Results, 'Overall behavioral changes'] The paper reports many significance tests across two behaviors, multiple pairwise contrasts, temporal trajectories, and exploratory heterogeneity analyses, but no adjustment for multiple comparisons is described. The pooled omnibus p=0.009 appears driven largely by electricity, while the hot-water contrast is p=0.087. Please clarify whether the two primary outcomes and the C-vs-T2 contrast were pre-specified in a registration or analysis plan, and report either a clear hypothesis hierarchy or multiplicity-adjusted p-values for the primary pair of outcomes and contrasts.
minor comments (4)
  1. [Abstract and Introduction] The abstract and Introduction describe electricity and hot-water use as 'objectively measured' and 'measured using dormitory meter readings and billing records,' but the Methods explicitly state that outcome data were self-reported via the chatbot. Please revise the wording to avoid implying independent metering.
  2. [Methods, 'Experiment design'] The baseline psychological profile is described as 'empirically validated in our vignette pilot study [73].' Since reference [73] is a self-citation, please provide validation details or external evidence for the scale in the main text or supplement.
  3. [Methods, 'Nudge effect estimation'] The ensemble ITE equation is typeset with a missing denominator or implicit 1/4 factor; please clarify the formula and define all terms explicitly, including the indexing of the meta-learners.
  4. [Results, 'Participant engagement and interaction'] The engagement-rate values reported in the text (57.1%, 58.2%, 69.7%) appear inconsistent with values visible in Figure 3d (which appear to show 42.8%, 32.8%, 32.2%). Please check whether the figure or text refers to a different metric, such as survival at a particular round, and label accordingly.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; minor self-citation for covariate validation does not drive the experimental contrast.

full rationale

The paper's central claim—that LLM-personalized nudges (T2) reduced electricity consumption by 0.56 kWh per room-day (p=0.014) relative to text-based feedback (C)—is a causal estimate from a randomized field experiment, not a quantity derived from a fitted equation, a definitional identity, or a prior publication by the same authors. The adjusted saving rates are standard summaries of model-predicted consumption and do not feed back into the treatment assignment or the outcome measurement. The content-analysis and engagement results are descriptive manipulation checks rather than inputs to the primary estimate. The only self-referential element is the Methods statement that the baseline pro-conservation psychological profile was 'empirically validated in our vignette pilot study [73]', which is a citation to the authors' own prior work. That covariate battery is used for adjustment and exploratory heterogeneity analyses, but the main causal contrast does not reduce to it, and the treatment effect would stand independently even if that covariate were omitted. Concerns about self-reported meter readings and possible differential under-reporting are measurement-validity threats, not circularity: they attack the reliability of the outcome data, not the derivation of the estimate from those data. Accordingly, no specific circular step is exhibited, and the paper's primary result should not be considered circular.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim depends on standard experimental-design assumptions plus the measurement assumption that self-reported meter readings reflect true consumption. The psychological profile covariate rests on the authors' own earlier pilot. There are no invented entities or fitted ad-hoc constants beyond the data-cleaning thresholds.

free parameters (1)
  • Participant exclusion criteria = >40% missing daily observations; extreme or unstable consumption patterns
    Researcher-defined thresholds determine the analytic sample (electricity n=169/233; hot water n=166/233). The exact definition of 'extreme or unstable' is not fully specified in the text, and this choice can influence treatment effect estimates. Lee bounds partly address attrition but do not eliminate the sensitivity.
assumptions (3)
  • domain assumption Random assignment of co-participant clusters yields exchangeability between arms
    Standard RCT assumption; co-participants were assigned jointly to reduce spillover, but other unmeasured social interactions could still violate no-interference.
  • domain assumption Self-reported meter readings are an unbiased proxy for actual consumption
    Outcome data were self-reported via the chatbot and only partially reconciled; systematic under-reporting in the LLM-personalized arm would bias the treatment effect.
  • ad hoc to paper The baseline psychological profile scale is a valid covariate
    Validation is cited to the authors' own vignette pilot study [73], with no external validation; this scale is used as a covariate in the main analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LLM-generated personalized nudges for improving pro-environmental behavior: Field evidence from resource conservation." pith.science (2026). https://pith.science/paper/5UKUGOYU

@misc{pith2026260403881,
  author       = {Pith},
  title        = {Pith review of: LLM-generated personalized nudges for improving pro-environmental behavior: Field evidence from resource conservation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5UKUGOYU}},
  note         = {Machine review of arXiv:2604.03881}
}
read the original abstract

Encouraging pro-environmental behavior remains a major challenge for sustainable cities. Conventional feedback nudges can show individuals how their current behavior compares with environmental goals but often provide limited guidance on what to do differently in daily life. This study examines whether supplementing weekly feedback on participants' behavior with LLM-generated personalized action suggestions improves pro-environmental behavior, using daily electricity and hot-water conservation as a case study. We developed an LLM agent that generated weekly conservation messages from participant profiles, recent consumption records, and prior interaction history, combining a usage report with personalized suggestions, behavioral-change scenarios, and estimated savings. The agent was evaluated in a three-arm randomized field experiment with 233 university residents in Beijing from November 2024 to January 2025. Participants received text-based nudges, image-enhanced nudges, or LLM-generated personalized nudges over five intervention rounds. Daily electricity use and shower hot-water use were measured using dormitory meter readings and billing records. Compared with text-based feedback, LLM-generated personalized nudges reduced electricity consumption by 0.56 kWh per room-day (p = 0.014), corresponding to an 18.3 percentage-point higher saving rate. Image-enhanced feedback alone showed no clear improvement. Hot-water savings followed the same direction but were smaller and less precisely estimated (9.8 percentage points, p = 0.087). Personalized nudges contained more planning, appliance-specific, and action-oriented language and were associated with more sustained, task-focused engagement. These findings offer a pathway for integrating generative AI into sustainable urban management.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

81 extracted references · 3 linked inside Pith

  1. [1]

    OECD Publishing, Paris, 2024

    Organisation for Economic Co-operation Development.LOGIC: Good Practice Principles for Mainstreaming Behavioural Public Policy. OECD Publishing, Paris, 2024

    2024

  2. [2]

    From evidence to delivery: an implementation-science blueprint for behavioural policy.Be- havioural Public Policy, page 1–12, 2025

    Giuseppe Alessandro Veltri. From evidence to delivery: an implementation-science blueprint for behavioural policy.Be- havioural Public Policy, page 1–12, 2025

    2025

  3. [3]

    Zimbardo and Michael R

    Philip G. Zimbardo and Michael R. Leippe.The Psychology of Attitude Change and Social Influence. Temple University Press, Philadelphia, illustrated edition edition, 1991

    1991

  4. [4]

    Goldberg, and Sander van der Linden

    Magnus Bergquist, Maximilian Thiel, Matthew H. Goldberg, and Sander van der Linden. Field interventions for climate change mitigation behaviors: A second-order meta-analysis.Proceed- ings of the National Academy of Sciences, 120(13):e2214851120, 2023

    2023

  5. [5]

    Milkman, Xiaomin Li, Angela L

    Anastasia Buyalskaya, Hung Ho, Katherine L. Milkman, Xiaomin Li, Angela L. Duckworth, and Colin Camerer. What can machine learning teach us about habit formation? evidence from exercise and hygiene.Proceedings of the National Academy of Sciences, 120(17):e2216115120, 2023

    2023

  6. [6]

    Michie, M

    S. Michie, M. M. van Stralen, and R. West. The behaviour change wheel: a new method for characterising and designing behaviour change interventions.Implement Sci, 6:42, 2011

    2011

  7. [7]

    Self- regulation and goal-directed behavior: A systematic literature review, public policy recommendations, and research agenda

    Soniya Billore, Tatiana Anisimova, and Demetris Vrontis. Self- regulation and goal-directed behavior: A systematic literature review, public policy recommendations, and research agenda. Journal of Business Research, 156:113435, 2023

    2023

  8. [8]

    Jamison, Berk Özler, Luca Parisotto, and Julius Dohbit Sama

    Susan Athey, Katy Bergstrom, Vitor Hadad, Julian C. Jamison, Berk Özler, Luca Parisotto, and Julius Dohbit Sama. Can per- sonalized digital counseling improve consumer search for mod- ern contraceptive methods?Science Advances, 9(40):eadg4420, 2023

    2023

Show all 81 references
  1. [9]

    N. K. Janz and M. H. Becker. The health belief model: a decade later.Health Educ Q, 11(1):1–47, 1984

    1984

  2. [10]

    E. A. Locke and G. P. Latham. Building a practically useful theory of goal setting and task motivation. a 35-year odyssey.Am Psychol, 57(9):705–17, 2002

    2002

  3. [11]

    Wood and D

    W. Wood and D. T. Neal. A new look at habits and the habit-goal interface.Psychol Rev, 114(4):843–63, 2007

    2007

  4. [12]

    R. M. Ryan and E. L. Deci. Self-determination theory and the facilitation of intrinsic motivation, social development, and well- being.Am Psychol, 55(1):68–78, 2000

    2000

  5. [13]

    Koch, Dan Mønster, and Julia Nafziger

    Alexander K. Koch, Dan Mønster, and Julia Nafziger. Spillover effects of reminder nudges in complex environments.Proceed- ings of the National Academy of Sciences, 121(17):e2322549121, 2024

    2024

  6. [14]

    Y . Luo, A. Li, D. Soman, and J. Zhao. A meta-analytic cognitive framework of nudge and sludge.R Soc Open Sci, 10(11):230053, 2023

    2023

  7. [15]

    Personalized nudging.Behavioural Public Policy, 6(1):150–159, 2022

    Stuart Mills. Personalized nudging.Behavioural Public Policy, 6(1):150–159, 2022

    2022

  8. [16]

    E. A. G. Hietbrink, C. Lansink, G. D. Laverman, M. M. R. V ollenbroek-Hutten, A. Middelweerd, and M. Tabak. Systematic review of dynamically tailored ehealth interventions targeting physical activity and healthy diet in chronic disease.npj Digital Medicine, 8(1):696, 2025

    2025

  9. [17]

    Can social information programs be more effective? the role of environmental identity for energy conservation.Journal of Environmental Economics and Management, 108:102467, 2021

    Jacopo Bonan, Cristina Cattaneo, Giovanna d’Adda, and Massimo Tavoni. Can social information programs be more effective? the role of environmental identity for energy conservation.Journal of Environmental Economics and Management, 108:102467, 2021

    2021

  10. [18]

    Behavioral barriers impede pro-environmental decision-making: Experimen- tal evidence from incentivized laboratory and vignette studies

    Roman Hoffmann, Georg Kanitsar, and Marcel Seifert. Behavioral barriers impede pro-environmental decision-making: Experimen- tal evidence from incentivized laboratory and vignette studies. Ecological Economics, 225:108347, 2024

    2024

  11. [19]

    Kraft, David Blazar, and Dylan Hogan

    Matthew A. Kraft, David Blazar, and Dylan Hogan. The effect of teacher coaching on instruction and achievement: A meta- analysis of the causal evidence.Review of Educational Research, 88(4):547–588, 2018

    2018

  12. [20]

    Instructional coaching personnel and program scalability.Education Finance and Policy, 19:1–64, 2023

    David Blazar, Doug McNamara, and Genine Blue. Instructional coaching personnel and program scalability.Education Finance and Policy, 19:1–64, 2023

    2023

  13. [21]

    Sara Mahdavi, Christopher Semturs, Juraj Gottweis, Joelle Barral, Katherine Chou, Greg S

    Tao Tu, Mike Schaekermann, Anil Palepu, Khaled Saab, Jan Freyberg, Ryutaro Tanno, Amy Wang, Brenna Li, Mohamed Amin, Yong Cheng, Elahe Vedadi, Nenad Tomasev, Shekoofeh Azizi, Karan Singhal, Le Hou, Albert Webson, Kavita Kulkarni, S. Sara Mahdavi, Christopher Semturs, Juraj Got...

    2025

  14. [22]

    Exploring the potential of llms as personalized assistants: Dataset, evaluation, and analysis

    Jisoo Mok, Ik-hwan Kim, Sangkwon Park, and Sungroh Yoon. Exploring the potential of llms as personalized assistants: Dataset, evaluation, and analysis. Proceedings of the 63rd Annual Meet- ing of the Association for Computational Linguistics (V olume 1: Long Papers), page 1021...

  15. [23]

    Enhancing beha vioral nudges with large language model-based iterative personalization15 Conversational planning for personal plans.arXiv preprint arXiv:2502.19500, 2025

    Konstantina Christakopoulou, Iris Qu, John Canny, Andrew Goodridge, Cj Adams, Minmin Chen, and Maja Matari ´c. Enhancing beha vioral nudges with large language model-based iterative personalization15 Conversational planning for personal plans.arXiv preprint arXiv:2502.19500, 2025

    2025 arXiv

  16. [24]

    Llm-based conversational agents for behaviour change support: A randomised controlled trial examining efficacy, safety, and the role of user behaviour

    Selina Meyer and David Elsweiler. Llm-based conversational agents for behaviour change support: A randomised controlled trial examining efficacy, safety, and the role of user behaviour. International Journal of Human-Computer Studies, 200:103514, 2025

    2025

  17. [25]

    Bean, Rebecca Elizabeth Payne, Guy Parsons, Han- nah Rose Kirk, Juan Ciro, Rafael Mosquera-Gómez, Sara Hin- capié M, Aruna S

    Andrew M. Bean, Rebecca Elizabeth Payne, Guy Parsons, Han- nah Rose Kirk, Juan Ciro, Rafael Mosquera-Gómez, Sara Hin- capié M, Aruna S. Ekanayaka, Lionel Tarassenko, Luc Rocher, and Adam Mahdi. Reliability of llms as medical assistants for the general public: a randomized prer...

    2026

  18. [26]

    Evaluation and mitigation of the limitations of large lan- guage models in clinical decision-making.Nature Medicine, 30(9):2613–2622, 2024

    Paul Hager, Friederike Jungmann, Robbie Holland, Kunal Bha- gat, Inga Hubrecht, Manuel Knauer, Jakob Vielhauer, Marcus Makowski, Rickmer Braren, Georgios Kaissis, and Daniel Rueck- ert. Evaluation and mitigation of the limitations of large lan- guage models in clinical decisio...

    2024

  19. [27]

    Argyle, Christopher A

    Lisa P. Argyle, Christopher A. Bail, Ethan C. Busby, Joshua R. Gubler, Thomas Howe, Christopher Rytting, Taylor Sorensen, and David Wingate. Leveraging ai for democratic discourse: Chat interventions can improve online political conversations at scale.Proceedings of the Nation...

    2023

  20. [28]

    Ocean- chat: The effect of virtual conversational ai agents on sustainable attitude and behavior change.arXiv preprint arXiv:2502.02863, 2025

    Pat Pataranutaporn, Alexander Doudkin, and Pattie Maes. Ocean- chat: The effect of virtual conversational ai agents on sustainable attitude and behavior change.arXiv preprint arXiv:2502.02863, 2025

    2025 arXiv

  21. [29]

    Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G

    Kobi Hackenburg, Ben M. Tappin, Luke Hewitt, Ed Saunders, Sid Black, Hause Lin, Catherine Fist, Helen Margetts, David G. Rand, and Christopher Summerfield. The levers of political persuasion with conversational artificial intelligence.Science, 390(6777):eaea3884, 2025

    2025

  22. [30]

    Hardy, Thomas L

    Rachit Dubey, Mathew D. Hardy, Thomas L. Griffiths, and Rahul Bhui. Ai-generated visuals of car-free us cities help improve sup- port for sustainable policies.Nature Sustainability, 7(4):399–403, 2024

    2024

  23. [31]

    Costello, Gordon Pennycook, and David G

    Thomas H. Costello, Gordon Pennycook, and David G. Rand. Durably reducing conspiracy beliefs through dialogues with ai. Science, 385(6714):eadq1814, 2024

    2024

  24. [32]

    Heinz, Daniel M

    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. Ja- cobson. Randomized trial of a generative ai chatbot for mental health treatment.NEJM AI, 2(4), 2025

    2025

  25. [33]

    Young, Francine C

    Andrew Allen, Allan H. Young, Francine C. Jellesma, Anton V orobev, Evgeniia Ivanova, Nikolay Babakov, Ani Gisnarian, and Lee Kannis-Dymand. An exploratory randomized controlled trial of an ai-enabled mental health intervention for generalized anxiety.Journal of Affective Diso...

    2026

  26. [34]

    Sinclair, Danielle Cosme, Kirsten Lydic, Diego A

    Alyssa H. Sinclair, Danielle Cosme, Kirsten Lydic, Diego A. Reinero, Michael E. Mann, and Emily B. Falk. Reply to dablander et al.: Identifying interventions that change intentions lays a valuable foundation for behavior change.Proceedings of the National Academy of Sciences, ...

    2025

  27. [35]

    To- ward achieving persistent behavior change in household water conservation.Proceedings of the National Academy of Sciences, 118(24):e2023014118, 2021

    Vivek Vivek, Deepak Malghan, and Kanchan Mukherjee. To- ward achieving persistent behavior change in household water conservation.Proceedings of the National Academy of Sciences, 118(24):e2023014118, 2021

    2021

  28. [36]

    Rafter, and Kira E

    Omar Isaac Asensio, Olga Churkina, Becky D. Rafter, and Kira E. O’Hare. Housing policies and energy efficiency spillovers in low and moderate income communities.Nature Sustainability, 7(5):590–601, 2024

    2024

  29. [37]

    Consumer purchases of energy-efficient appliances: A systematic literature review and research agenda

    Hoang Viet Nguyen, Bao Ngoc Le, Weng Marc Lim, Thac Dang- Van, and Ninh Nguyen. Consumer purchases of energy-efficient appliances: A systematic literature review and research agenda. Energy Efficiency, 18(4):29, 2025

    2025

  30. [38]

    One thing leads to another: Evidence on the scope and persistence of be- havioral spillovers.Journal of Public Economics, 236:105166, 2024

    Alexander Goetz, Harald Mayr, and Renate Schubert. One thing leads to another: Evidence on the scope and persistence of be- havioral spillovers.Journal of Public Economics, 236:105166, 2024

    2024

  31. [39]

    Jin Han Kim and Michael Kaemingk. Persisting effects of social norm feedback letters in reducing household electricity usage in post-soviet eastern europe: A randomized controlled trial.Journal of Economic Behavior&Organization, 191:153–161, 2021

    2021

  32. [40]

    Smart metering and systems to support a conscious use of water and electricity.Energy, 45(1):528–540, 2012

    Carlos Alberto Fróes Lima and José Ricardo Portillo Navas. Smart metering and systems to support a conscious use of water and electricity.Energy, 45(1):528–540, 2012

    2012

  33. [41]

    Hoppen, Rieke M

    Thole H. Hoppen, Rieke M. Cuno, Janna Nelson, Frederike Lem- mel, Pascal Schlechter, and Nexhmedin Morina. Meta-analysis of randomized controlled trials examining social comparison as a behaviour change technique across the behavioural sciences. Nature Human Behaviour, 2025

    2025

  34. [42]

    Rank ordering information on energy use can promote energy-conserving behavior in china.Communications Earth&Environment, 6(1):947, 2025

    Wenbo Li, Xiu Cheng, Ruyin Long, Jiaxin Huang, and Chengjiang Li. Rank ordering information on energy use can promote energy-conserving behavior in china.Communications Earth&Environment, 6(1):947, 2025

    2025

  35. [43]

    Motivating metrics for household water-use feedback.Resources, Conserva- tion and Recycling, 103:29–46, 2015

    Ariane Liu, Damien Giurco, and Pierre Mukheibir. Motivating metrics for household water-use feedback.Resources, Conserva- tion and Recycling, 103:29–46, 2015

    2015

  36. [44]

    The design of an energy and water advice programme for low-income households.Energy and Buildings, 110:426–434, 2016

    Marina Laskari, Stavroula Karatasou, and Mat Santamouris. The design of an energy and water advice programme for low-income households.Energy and Buildings, 110:426–434, 2016

    2016

  37. [45]

    Converging conventions of comfort, cleanliness and convenience.Journal of Consumer Policy, 26(4):395–418, 2003

    Elizabeth Shove. Converging conventions of comfort, cleanliness and convenience.Journal of Consumer Policy, 26(4):395–418, 2003

    2003

  38. [46]

    Dynamic thermal responses and showering thermal comfort under different conditions.Building and Environment, 237:110322, 2023

    Maohui Luo, Shuang Xu, Yin Tang, Hang Yu, Xiang Zhou, and Zong Chen. Dynamic thermal responses and showering thermal comfort under different conditions.Building and Environment, 237:110322, 2023

    2023

  39. [47]

    turning offthe lights

    Daniel C. Lundberg, Janine A. Tang, and Shahzeen Z. Attari. Easy but not effective: Why “turning offthe lights” remains a salient energy conserving behaviour in the united states.Energy Research&Social Science, 58:101257, 2019

    2019

  40. [48]

    Nudge versus boost: How coherent are policy and theory?Minds and Machines, 26(1):149–183, 2016

    Till Grüne-Yanoffand Ralph Hertwig. Nudge versus boost: How coherent are policy and theory?Minds and Machines, 26(1):149–183, 2016

    2016

  41. [49]

    Boosts vs nudges: perceived effectiveness and criticism shape preferences for sustainable be- havioural policies.Energy Policy, 209:114953, 2026

    Yavor Paunov and Till Grüne-Yanoff. Boosts vs nudges: perceived effectiveness and criticism shape preferences for sustainable be- havioural policies.Energy Policy, 209:114953, 2026

    2026

  42. [50]

    Grelle and W

    S. Grelle and W. Hofmann. When and why do people ac- cept public-policy interventions? an integrative public-policy- acceptance framework.Perspect Psychol Sci, 19(1):258–279, 2024

    2024

  43. [51]

    Nudging: Progress to date and future directions.Organizational Behavior and Human Deci- sion Processes, 161:3–19, 2020

    John Beshears and Harry Kosowsky. Nudging: Progress to date and future directions.Organizational Behavior and Human Deci- sion Processes, 161:3–19, 2020

    2020

  44. [52]

    Patterson and William Skimmyhorn

    Richard W. Patterson and William Skimmyhorn. How do behav- ioral approaches to increase savings compare? evidence from multiple interventions in the u.s. army.Management Science, 2025. Enhancing beha vioral nudges with large language model-based iterative personalization16

    2025

  45. [53]

    Han, Lily Roh, Naveen Raja, Sitaram Vangala, Hardikkumar Modi, Shital Pandya, Michael Sloyan, and Daniel M

    Hengchen Dai, Silvia Saccardo, Maria A. Han, Lily Roh, Naveen Raja, Sitaram Vangala, Hardikkumar Modi, Shital Pandya, Michael Sloyan, and Daniel M. Croymans. Behavioural nudges in- crease covid-19 vaccinations.Nature, 597(7876):404–409, 2021

    2021

  46. [54]

    Routledge, 2023

    Stefan Drews and Jeroen CJM van den Bergh.A critical assess- ment of the effectiveness of low-carbon nudges, page 268–284. Routledge, 2023

    2023

  47. [55]

    Nudging towards sustainability: a comprehensive review of behavioral ap- proaches to eco-friendly choice.Discover Sustainability, 5(1):444, 2024

    Babak Amiri, Amirhossein Jafarian, and Zahra Abdi. Nudging towards sustainability: a comprehensive review of behavioral ap- proaches to eco-friendly choice.Discover Sustainability, 5(1):444, 2024

    2024

  48. [56]

    Bibhas Chakraborty and Susan A. Murphy. Dynamic treatment regimes.Annual Review of Statistics and Its Application, 1(V ol- ume 1, 2014):447–464, 2014

    2014

  49. [57]

    Bridging innovation and equity: Advancing public health through just-in-time adaptive interventions.Annual Review of Public Health, 46(V olume 46, 2025):43–68, 2025

    Nur Hani Zainal, Xueqing Liu, Utek Leong, Xiaoxi Yan, and Bibhas Chakraborty. Bridging innovation and equity: Advancing public health through just-in-time adaptive interventions.Annual Review of Public Health, 46(V olume 46, 2025):43–68, 2025

    2025

  50. [58]

    Nahum-Shani, S

    I. Nahum-Shani, S. N. Smith, B. J. Spring, L. M. Collins, K. Witkiewitz, A. Tewari, and S. A. Murphy. Just-in-time adap- tive interventions (jitais) in mobile health: Key components and design principles for ongoing health behavior support.Ann Behav Med, 52(6):446–462, 2018

    2018

  51. [59]

    N. Bol, E. S. Smit, and M. L. A. Lustria. Tailored health commu- nication: Opportunities and challenges in the digital era.Digit Health, 6:2055207620958913, 2020

    2020

  52. [60]

    S. A. Lee and J. H. Park. Systematic review and meta analysis of standalone digital behavior change interventions on physical activity.NPJ Digit Med, 8(1):436, 2025

    2025

  53. [61]

    The development and use of ai chatbots for health behavior change: Scoping review.J Med Internet Res, 28:e79677, 2026

    Lingyi Fu, Ryan Burns, Yuhuan Xie, Jincheng Shen, Shandian Zhe, Paul Estabrooks, and Yang Bai. The development and use of ai chatbots for health behavior change: Scoping review.J Med Internet Res, 28:e79677, 2026

    2026

  54. [62]

    Karine Karine and Benjamin M. Marlin. Enhancing adaptive behavioral interventions with llm inference from participant de- scribed states, 2025

    2025

  55. [63]

    Rothkopf, and Kristian Kersting

    Patrick Schramowski, Cigdem Turan, Nico Andersen, Con- stantin A. Rothkopf, and Kristian Kersting. Large pre-trained language models contain human-like biases of what is right and wrong to do.Nature Machine Intelligence, 4(3):258–268, 2022

    2022

  56. [64]

    Whose opinions do language models reflect?, 2023

    Shibani Santurkar, Esin Durmus, Faisal Ladhak, Cinoo Lee, Percy Liang, and Tatsunori Hashimoto. Whose opinions do language models reflect?, 2023

    2023

  57. [65]

    A large-scale replication of scenario-based experiments in psychology and management using large language models.Nature Computational Science, 5(8):627–634, 2025

    Ziyan Cui, Ning Li, and Huaikang Zhou. A large-scale replication of scenario-based experiments in psychology and management using large language models.Nature Computational Science, 5(8):627–634, 2025

    2025

  58. [66]

    Talarico

    Jennifer M. Talarico. The merits of undergraduate student samples. Nature Reviews Psychology, 2025

    2025

  59. [67]

    Retrieval-augmented generation for knowledge-intensive nlp tasks, 2020

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. Retrieval-augmented generation for knowledge-intensive nlp tasks, 2020

    2020

  60. [68]

    Chi, Quoc V

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V . Le, and Denny Zhou. Chain-of-thought prompting elicits reasoning in large language models, 2022

    2022

  61. [69]

    L. A. Palinkas, S. M. Horwitz, C. A. Green, J. P. Wisdom, N. Duan, and K. Hoagwood. Purposeful sampling for qualitative data collection and analysis in mixed method implementation research.Adm Policy Ment Health, 42(5):533–44, 2015

    2015

  62. [70]

    Ortiga, Jue Wang, and Meng-Hsuan Chou

    Yasmin Y . Ortiga, Jue Wang, and Meng-Hsuan Chou. Interna- tional academic or citizen scholar?: Singaporean returnees in the global city.Cities, 137:104327, 2023

    2023

  63. [71]

    Petersen, Cynthia M

    John E. Petersen, Cynthia M. Frantz, Md Rumi Shammin, Tess M. Yanisch, Evan Tincknell, and Noel Myers. Electricity and water conservation on college and university campuses in response to national competitions among dormitories: Quantifying relation- ships between behavior, co...

    2015

  64. [72]

    Social foundations of thought and action.Engle- wood Cliffs, NJ, 1986(23-28):2, 1986

    Albert Bandura. Social foundations of thought and action.Engle- wood Cliffs, NJ, 1986(23-28):2, 1986

    1986

  65. [73]

    Potential of large language model-powered nudges for promoting daily water and energy conservation.arXiv preprint arXiv:2503.11531, 2025

    Zonghan Li, Song Tong, Yi Liu, Kaiping Peng, and Chunyan Wang. Potential of large language model-powered nudges for promoting daily water and energy conservation.arXiv preprint arXiv:2503.11531, 2025

    2025 arXiv

  66. [74]

    quanteda: An r package for the quantitative analysis of textual data.Journal of Open Source Software, 3:774, 2018

    Kenneth Benoit, Kohei Watanabe, Haiyan Wang, Paul Nulty, Adam Obeng, Stefan Müller, and Akitaka Matsuo. quanteda: An r package for the quantitative analysis of textual data.Journal of Open Source Software, 3:774, 2018

    2018

  67. [75]

    Keyword- assisted topic models.American Journal of Political Science, 68(2):730–750, 2024

    Shusei Eshima, Kosuke Imai, and Tomoya Sasaki. Keyword- assisted topic models.American Journal of Political Science, 68(2):730–750, 2024

    2024

  68. [76]

    Künzel, Jasjeet S

    Sören R. Künzel, Jasjeet S. Sekhon, Peter J. Bickel, and Bin Yu. Metalearners for estimating heterogeneous treatment effects using machine learning.Proceedings of the National Academy of Sciences, 116(10):4156–4165, 2019

    2019

  69. [77]

    Edward H. Kennedy. Towards optimal doubly robust estimation of heterogeneous causal effects.Electronic Journal of Statistics, 2020

    2020

  70. [78]

    Long-term water conservation is fostered by smart meter-based feedback and digital user engagement.npj Clean Water, 4(1):29, 2021

    Andrea Cominola, Matteo Giuliani, Andrea Castelletti, Piero Fraternali, Sergio Luis Herrera Gonzalez, Joan Carles Guardiola Herrero, Jasminko Novak, and Andrea Emilio Rizzoli. Long-term water conservation is fostered by smart meter-based feedback and digital user engagement.np...

    2021

  71. [79]

    Cambridge University Press, 2022

    Steven L Brunton and J Nathan Kutz.Data-driven science and engineering: Machine learning, dynamical systems, and control. Cambridge University Press, 2022

    2022

  72. [80]

    A comparison study on similarity and dissimilarity measures in clustering continuous data.PloS one, 10(12):e0144059, 2015

    Ali Seyed Shirkhorshidi, Saeed Aghabozorgi, and Teh Ying Wah. A comparison study on similarity and dissimilarity measures in clustering continuous data.PloS one, 10(12):e0144059, 2015

    2015

  73. [81]

    Xgboost: A scalable tree boosting system, 2016

    Tianqi Chen and Carlos Guestrin. Xgboost: A scalable tree boosting system, 2016

    2016

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.