Pith. sign in

REVIEW 3 major objections 5 minor 55 references

GPTFootprint: Increasing Consumer Awareness of the Environmental Impacts of LLMs

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read GPTFootprint increases ChatGPT users' environmental awareness, but leaves total query volume roughly unchanged.

desk verdict Honest, well-designed little systems paper whose central awareness claim outstrips the evidence; worth engaging, but the evaluation needs a redesign and the server logs should be used. read the letter →

arxiv 2505.24107 v1 pith:ZPD6APYI submitted 2025-05-30 cs.HC

Nora Graves , Vitus Larrieu , Yingyue Trace Zhang , Joanne Peng , Varun Nagaraj Rao , Yuhan Liu , Andrés Monroy-Hernández This is my paper
classification cs.HC
keywords eco-feedbacksystemsenvironmentalawarenesslargelanguagemodelsbehaviorchangeChatGPTbrowserextensionenergyconsumptionwater
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 introduces GPTFootprint, a browser extension that shows ChatGPT users live estimates of the energy and water their queries consume, and asks whether this feedback increases awareness and reduces usage. In a week-long study with nine regular ChatGPT users, participants reported higher awareness of LLM environmental impacts and appreciation for human-scale metrics, but total queries did not decline. The paper concludes that awareness-based individual interventions can educate users, yet face a hard barrier: ChatGPT's practical utility outweighs the felt environmental cost, so usage reduction requires more than information display.

What carries the argument

The intervention's load-bearing component is the Eco Score: a 0-to-100 grade-like metric that falls by 7 to 13 points per query depending on the pause length since the previous query, with rapid successive queries costing the most, and that recovers 1 point every 20 minutes so an efficient day resets overnight. The score sits in an always-visible side panel that also converts cumulative energy and water use into changing human-scale icons, and a popup appears every seven queries to invite a break. The Eco Score gamifies the feedback to avoid the guilt and stress observed in pilot versions, and the pause-based deduction is designed to reward users who seek alternatives rather than querying ChatGPT in quick succession.

What would settle it

Run a controlled trial with a larger, diverse sample in which one group gets GPTFootprint, a second gets a placebo extension showing only query counts without environmental metrics, and a third gets no extension; objectively log ChatGPT queries via the extension and compare pre- and post-trial awareness surveys. If the placebo group shows the same awareness increase, or if the GPTFootprint group's usage change is statistically indistinguishable from control, the claim that the metrics caused the awareness gain would be falsified.

Watch

Extended reading notes

Core claim

GPTFootprint computes each user's running energy and water totals from fixed per-query averages of 2.9Wh and 16.9mL, renders them as an Eco Score plus pictogram-based units such as lightbulb hours and cups of water, and shows a break-reminder popup after seven queries. In the user study, average self-rated learning was 4.11 out of 5, concern about the environmental impact of queries rose from 3.0 to 3.55 out of 5, and all participants reported new awareness, with strong emotional reactions including guilt. However, total queries across participants increased by 18.584% during the trial, and participants attributed continued use to task necessity; the popup only sometimes produced long pauses. The paper's central claim is therefore that GPTFootprint increases awareness of environmental impact but has limited success in decreasing ChatGPT usage.

Load-bearing premise

The study's central result depends on the assumption that self-reported Likert-scale awareness and participants' own ChatGPT export counts from just nine users, with no control group, reflect genuine changes rather than participants telling the researchers what they expected to hear or simply regressing to the mean.

Editorial extensions

If this is right

  • If awareness is the goal, GPTFootprint-style displays work: participants consistently reported new understanding and rated the tool 4.22 out of 5 on enjoyment.
  • Awareness alone will not reliably cut queries: most users judged ChatGPT's utility as greater than the environmental cost, so interventions need to change the cost-benefit balance or the task context.
  • The break popup behaves as a nudge rather than a blocker: most popups were followed by another query within 10 minutes, but a minority preceded gaps ranging from 24 minutes to over 24 hours, suggesting it helps users who are already considering stopping.
  • Privacy-preserving estimation is feasible: using fixed per-query averages and local tracking avoids reading query content while still conveying a personalized environmental impact.

Reading between the lines

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

  • A longer-term deployment might show different usage effects than the one-week trial: the paper's own reasoning suggests cumulative resource consumption could weigh more heavily over months, and the Eco Score's overnight recovery makes daily resets forgiving, so a multi-month study with a control group would test this directly.
  • The Eco Score's pause-based penalty assumes that longer pauses mean users are seeking alternative resources, but the paper notes this proxy is not validated; instrumenting search-engine or other-tool use could confirm whether longer pauses actually indicate substitution.
  • Because the sample was nine university students, the awareness effect may not transfer to populations with different prior knowledge of LLMs or different stakes such as job-related use, and broader deployment could reveal whether guilt is replaced by other emotional or practical responses.
  • Making individual footprints visible risks shifting blame onto end users while companies control model efficiency, a design-ethics tension the paper raises but does not resolve; future work could embed corporate-responsibility messaging alongside personal feedback.
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 / 5 minor

Summary. GPTFootprint is a Chrome extension that overlays the ChatGPT interface with a side panel showing per-query energy and water consumption (2.9 Wh and 16.9 mL per query), a gamified Eco Score, and a popup that appears after a query limit. The authors describe a seven-day, nine-participant single-arm user study in which participants self-reported increased awareness of the environmental impact of LLMs, while ChatGPT query counts, computed by a participant-run Colab notebook, did not decrease overall. The paper concludes that individual-level awareness interventions can be effective at increasing awareness but have limited success at reducing usage, and it offers qualitative themes about utility and personal responsibility.

Significance. The system design is thoughtful and well-documented: it is privacy-preserving (no query content is accessed), uses human-scale pictograms and a transparent Eco Score algorithm, and the appendices include the survey instruments and a query-counting Colab notebook. If the results hold, the qualitative findings about utility and perceived limits of individual responsibility would be useful to the sustainable-AI and eco-feedback communities. However, the central awareness claim is currently supported only by self-reports from a small, uncontrolled sample, and the usage data are not validated against the objectively logged server events; the evidence as presented is not yet sufficient to establish the causal claims in the abstract.

major comments (3)
  1. [§5.1, Table 2] The abstract and conclusion claim that GPTFootprint 'increases people's awareness of environmental impact,' but no survey item directly measures awareness. The only pre/post indicator, 'Care about the Environmental Impact of Queries,' moves from 3.00 to 3.55 on a 5-point scale with n=9, and no significance test or effect size is reported. The other supportive item, 'Learned More' (mean 4.11), asks participants to rate their own learning after using the intervention, which is vulnerable to demand characteristics. In a single-arm study in which participants were informed of the study's purpose, these results cannot rule out regression to the mean or social desirability. The claim should be narrowed to 'participants reported increased awareness' or supported by a validated awareness measure with a control group and inferential statistics.
  2. [§5.3, Table 1 vs §3.3, Appendix E] The usage conclusion rests on query counts derived from a Colab notebook that each participant ran on an exported conversations.json file. However, Section 3.3 states that the extension itself logged every query event (user ID, date/time, event type) to a private spreadsheet during the user studies. The paper never compares the participant-processed counts with these server logs. This validation is essential: if the notebook overcounts (for example, by counting all user-authored messages rather than completed queries) or mishandles the UTC date conversion, the reported 18.584% increase would be an artifact. The authors should report the server-log counts and reconcile any discrepancies.
  3. [§4 and §7] The limitations section acknowledges the homogeneous college-student sample but does not mention the absence of a control group or the reliance on self-report outcome measures. Because the central claim is causal ('increases awareness'), the single-arm design and the demand-characteristics risk should be explicitly discussed in the limitations. Adding a sentence about these threats and how the authors attempted to mitigate them would make the framing more honest and would help readers calibrate the strength of the conclusions.
minor comments (5)
  1. [§3.2] The sentence 'the full algorithm appears in Appendix ??' contains an unresolved placeholder; it should refer to Appendix B, where the Eco Score algorithm is actually given.
  2. [References] Reference [32] is incomplete: it reads 'Title of the paper as per the source (replace this placeholder)' and 'Proceedings of the ACM Conference/Journal Name (replace this placeholder).' The authors should replace it with the full citation for the Luccioni and Strubell work (for example, the FAccT 2024 paper 'Power Hungry Processing: Watts Driving the Cost of AI Deployment?').
  3. [Table 1] The red/green cell coloring used to indicate increases and decreases in query counts is not legible in a monochrome or print version; please add explicit '+' and '−' signs or arrows to each numeric cell.
  4. [§5.3] The phrase 'These user reactions indicate of a key limitation' has a grammatical error; it should be 'indicate a key limitation.'
  5. [§3.3.1 and Appendix H] The 'Read More' document is described as providing further information, but its content is only shown as an appendix image; a short summary of what the document says would help readers understand what information participants received.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the awareness and usage claims rest on independent survey and log data, not on the Eco Score parameters that the paper itself calibrated.

full rationale

GPTFootprint's central claims are empirical findings from a nine-participant week-long study, not mathematical derivations from fitted quantities. The Eco Score algorithm is a design artifact: its penalty tiers and 20-minute recovery rate are hand-chosen and pilot-calibrated to yield a score of 76 for an average user, but none of the reported outcomes—self-reported awareness gains, Likert responses, or before/after ChatGPT query counts—are computed from the Eco Score. The per-query energy (2.9 Wh) and water (16.9 mL) constants are taken from external sources (IEA; Li et al.), and the paper makes no claim to have derived those constants from its own data. The usage counts come from a participant-run Colab notebook over ChatGPT export data, not from the extension's Eco Score or from the paper's fitted parameters; the server logs (Appendix I) are an independent data source that the paper does not use to construct the reported result. No load-bearing step reduces, by construction or by self-citation, to its own inputs; the limitations noted by the authors (small sample, single arm, self-report) concern statistical validity, not circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The paper's intervention depends on externally sourced per-query resource numbers, a heuristic query detector, and self-report measurements. The Eco Score is a fully authored design artifact with tuned parameters. No circular, self-deriving claim appears in the evaluation.

free parameters (3)
  • Eco Score query deductions = 7 to 13 points per query depending on pause length
    Hand-tuned thresholds in Algorithm 1 (Appendix B) to make an 'average' user see a score near 76, just below 'good', with no derivation from independent theory.
  • Eco Score recovery rate = 1 point per 20 minutes
    Chosen so an efficient user with 6 queries an hour apart returns to 100 overnight (Section 3.2); the value is calibrated to the pilot average, not to an external outcome.
  • Popup query limit = 7 queries
    Increased from 3 after pilot feedback; this is a design parameter that directly sets how often the break prompt appears (Section 3.3.2).
assumptions (4)
  • domain assumption Per-query averages of 2.9 Wh energy and 16.9 mL water apply to every user query.
    Section 3.1 adopts these values from prior work (IEA; Li et al.) and applies them uniformly, despite the paper noting that query size and model affect real consumption.
  • domain assumption The extension's status-code detection (HTTP 200 to the ChatGPT backend, excluding 'init' and 'implicit' requests) reliably counts user queries.
    Section 3.3 states the design; false positives or missed requests would bias the displayed metrics and the logged usage data.
  • domain assumption The ChatGPT export-based Colab notebook counts the same queries as the extension's detector.
    Section 4 and Appendix E rely on participants running the notebook and reporting counts; the detection logic differs from the extension's, and the pre-trial window hinges on a self-reported download date.
  • domain assumption Self-reported Likert ratings and open-ended responses reflect genuine changes in awareness rather than demand characteristics.
    Section 5.1 analyzes awareness as an outcome of the intervention with no control condition or validated psychometric scale.
invented entities (1)
  • Eco Score
    purpose: A 0-100 gamified score that summarizes the environmental cost of a user's query pattern, with a deteriorating background image as the score drops.
    The score is an interface construct built from hand-tuned penalties and recovery; there is no external validation that it reflects actual environmental efficiency or that changing it causes behavior change.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GPTFootprint: Increasing Consumer Awareness of the Environmental Impacts of LLMs." pith.science (2026). https://pith.science/paper/ZPD6APYI

@misc{pith2026250524107,
  author       = {Pith},
  title        = {Pith review of: GPTFootprint: Increasing Consumer Awareness of the Environmental Impacts of LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZPD6APYI}},
  note         = {Machine review of arXiv:2505.24107}
}
read the original abstract

With the growth of AI, researchers are studying how to mitigate its environmental impact, primarily by proposing policy changes and increasing awareness among developers. However, research on AI end users is limited. Therefore, we introduce GPTFootprint, a browser extension that aims to increase consumer awareness of the significant water and energy consumption of LLMs, and reduce unnecessary LLM usage. GPTFootprint displays a dynamically updating visualization of the resources individual users consume through their ChatGPT queries. After a user reaches a set query limit, a popup prompts them to take a break from ChatGPT. In a week-long user study, we found that GPTFootprint increases people's awareness of environmental impact, but has limited success in decreasing ChatGPT usage. This research demonstrates the potential for individual-level interventions to contribute to the broader goal of sustainable AI usage, and provides insights into the effectiveness of awareness-based behavior modification strategies in the context of LLMs.

Figures

Figures reproduced from arXiv: 2505.24107 by the authors.

Figure 1
Figure 1. GPTFootprint displays the amount of energy and water consumed in user-friendly terms and with pictogram icons (hours powering a light bulb and cups of water) in a side panel on top of the ChatGPT interface. In addition, there is an Eco Score, which can range from 0 to 100. ∗All three authors contributed equally to this research. Permission to make digital or hard copies of part or all of this work for personal or cl… view at source ↗
Figure 2
Figure 2. System Display. (a) The side panel is always visible on screen, and displays an updating display of energy and water [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Study Participant Pre-Survey [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Study Participant Post-Survey [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Study Participant Post-Survey (cont.) [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 34 canonical work pages

  1. [32]

    Sasha Luccioni and Emma Strubell. 2023. Title of the paper as per the source (replace this placeholder). In Proceedings of the ACM Conference/Journal Name (replace this placeholder). https://doi.org/10.1145/3630106.3658542

    arXiv 2023

  2. [1]

    [n. d.]. Generational Growth: AI, Data Centers and the Coming US Power De- mand Surge. https://www.goldmansachs.com/pdfs/insights/pages/generational- growth-ai-data-centers-and-the-coming-us-power-surge/report.pdf

    work page

  3. [2]

    [n. d.]. Understanding Your Grades and Transcript. https://baruch- undergraduate.catalog.cuny.edu/policies-and-procedures/understanding-your- grades-and-transcript

    work page

  4. [3]

    How Much Sleep Is Enough? https://www.nhlbi.nih.gov/health/sleep/how- much-sleep#:~:text=Experts%20recommend%20that%20adults%20sleep,or% 20more%20hours%20a%20night

    2022. How Much Sleep Is Enough? https://www.nhlbi.nih.gov/health/sleep/how- much-sleep#:~:text=Experts%20recommend%20that%20adults%20sleep,or% 20more%20hours%20a%20night

    work page 2022

  5. [4]

    Ebina Justin M A and Manu Melwin Joy. 2024. Gamification, in- trinsic motivation, and task performance of employees: the moderat- ing role of goal difficulty. Behaviour & Information Technology 43, 16 (2024), 3993–4015. https://doi.org/10.1080/0144929X.2023.2297280 arXiv:https://doi.org/10.1080/0144929X.2023.2297280

    work page arXiv 2024

  6. [5]

    Environmental Protection Agency

    U.S. Environmental Protection Agency. 2025. How the Energy Star Score is Calculated. https://www.energystar.gov/buildings/benchmark/understand- metrics/how-score-calculated Accessed: 2025-01-21

    work page 2025

  7. [6]

    Ernesto Arroyo, Leonardo Bonanni, and Ted Selker. 2005. Waterbot: exploring feedback and persuasive techniques at the sink. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (Portland, Oregon, USA) (CHI ’05). Association for Computing Machinery, New York, NY, USA, 631–639. https://doi.org/10.1145/1054972.1055059

    arXiv 2005

  8. [7]

    Angels Art. 2025. Lightbulb Icon. https://thenounproject.com/icon/lightbulb- 3194358/. Accessed: 2025-01-23

    work page 2025

Show all 55 references
  1. [8]

    Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

    Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (Virtual Event, Canada) (FAcc...

    2021 doi

  2. [9]

    Ivo Blohm and Jan Marco Leimeister. 2013. Gamification. Bus. Inf. Syst. Eng. 5, 4 (Aug. 2013), 275–278

    2013

  3. [10]

    Dena Bravata, Crystal Smith-Spangler, Vandana Sundaram, Allison Gienger, Nancy Lin, Robyn Lewis, Christopher Stave, Ingram Olkin, and John Sirard

  4. [11]

    McDonald, and James A

    Sunny Consolvo, David W. McDonald, and James A. Landay. 2009. Theory-driven design strategies for technologies that support behavior change in everyday life. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (Boston, MA, USA) (CHI ’09). Association ...

    2009

  5. [12]

    Ben Cottier, Robi Rahman, Loredana Fattorini, Nestor Maslej, and David Owen

  6. [13]

    Ben Davis. 2025. Water Glass Icon. https://thenounproject.com/icon/water-glass- 1190124/. Accessed: 2025-01-23

    2025

  7. [14]

    Matheus Eduardo. 2024. ChatGPT Question Couunt. Chrome Extension Store (2024). https://chromewebstore.google.com/detail/chatgpt-question-count/ naokkoogmjjhnehoadkmpicliffbjllc

    2024

  8. [15]

    Jon Froehlich, Tawanna Dillahunt, Predrag Klasnja, Jennifer Mankoff, Sunny Consolvo, Beverly Harrison, and James Landay. 2009. UbiGreen: Investigating a Mobile Tool for Tracking and Supporting Green Transportation Habits. Proc. CHI 2009, 1043–1052. https://doi.org/10.1145/1518...

    2009

  9. [16]

    Jon Froehlich, Leah Findlater, and James Landay. 2010. The design of eco-feedback technology. In Proceedings of the SIGCHI Conference on Human Factors in Comput- ing Systems (Atlanta, Georgia, USA)(CHI ’10). Association for Computing Machin- ery, New York, NY, USA, 1999–2008. ...

    2010

  10. [17]

    Jon Froehlich, Leah Findlater, Marilyn Ostergren, Solai Ramanathan, Josh Pe- terson, Inness Wragg, Eric Larson, Fabia Fu, Mazhengmin Bai, Shwetak Pa- tel, and James A. Landay. 2012. The design and evaluation of prototype eco- feedback displays for fixture-level water usage dat...

    2012

  11. [18]

    Greenhill. 2025. Hot Tub Icon. https://thenounproject.com/icon/hot-tub- 6976573/. Accessed: 2025-01-23

    2025

  12. [19]

    Carlos Gómez-Rodríguez and Paul Williams. 2023. A Confederacy of Models: a Comprehensive Evaluation of LLMs on Creative Writing. https://doi.org/10. 48550/arXiv.2310.08433 arXiv:2310.08433 [cs]

    2023 doi arXiv

  13. [20]

    Tom Hargreaves. 2017. Beyond Energy Feedback. In Building Research & Infor- mation, Vol. 46. 332–342. https://doi.org/10.1080/09613218.2017.1356140

    2017

  14. [21]

    Iea. [n. d.]. Electricity 2024 – analysis. https://www.iea.org/reports/electricity- 2024

    2024

  15. [22]

    Rishee Jain, John Taylor, and Patricia Culligan. 2013. Investigating the impact eco-feedback information representation has on building occupant energy con- sumption behavior and savings. Energy and Buildings 64 (09 2013), 408–414. https://doi.org/10.1016/j.enbuild.2013.05.011

    2013 doi

  16. [23]

    Kaack, Priya L

    Lynn H. Kaack, Priya L. Donti, Emma Strubell, George Kamiya, Felix Creutzig, and David Rolnick. 2022. Aligning artificial intelligence with climate change mitigation. Nature Climate Change 12, 6 (June 2022), 518–527. https://doi.org/10. 1038/s41558-022-01377-7

    2022

  17. [24]

    Jaejeung Kim, Hayoung Jung, Minsam Ko, and Uichin Lee. 2019. GoalKeeper: Exploring Interaction Lockout Mechanisms for Regulating Smartphone Use. Proc. ACM Interact. Mob. Wearable Ubiquitous Technol. 3, 1, Article 16 (March 2019), 29 pages. https://doi.org/10.1145/3314403

    2019 doi

  18. [25]

    Jaejeung Kim, Joonyoung Park, Hyunsoo Lee, Minsam Ko, and Uichin Lee. 2019. LocknType: Lockout Task Intervention for Discouraging Smartphone App Use. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (Glasgow, Scotland Uk) (CHI ’19). Association f...

    2019

  19. [26]

    Alexandre Lacoste, Alexandra Luccioni, Victor Schmidt, and Thomas Dan- dres. 2019. Quantifying the Carbon Emissions of Machine Learning. arXiv:1910.09700 [cs.CY] https://arxiv.org/abs/1910.09700

    2019 arXiv

  20. [27]

    Islam, and Shaolei Ren

    Pengfei Li, Jianyi Yang, Mohammad A. Islam, and Shaolei Ren. 2023. Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models. arXiv:2304.03271 [cs.LG] https://arxiv.org/abs/2304.03271

    2023 arXiv

  21. [28]

    Feng Lin, Dong Jae Kim, Tse-Husn, and Chen. 2024. SOEN-101: Code Generation by Emulating Software Process Models Using Large Language Model Agents. https://doi.org/10.48550/arXiv.2403.15852 arXiv:2403.15852 [cs]

    2024 doi arXiv

  22. [29]

    Tao Lu, Hongxiao Zheng, Tianying Zhang, Xuhai “Orson” Xu, and Anhong Guo. 2024. InteractOut: Leveraging Interaction Proxies as Input Manipulation Strategies for Reducing Smartphone Overuse. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolu...

    2024

  23. [30]

    Pin Luarn, Chiao-Chieh Chen, and Chiu Yu-Ping. 2023. Enhancing intrinsic learning motivation through gamification: a self-determination theory per- spective. The International Journal of Information and Learning Technology 40, 5 (2023), 413–424. https://login.ezproxy.princeton...

    2023

  24. [31]

    Alexandra Sasha Luccioni, Sylvain Viguier, and Anne-Laure Ligozat. 2022. Es- timating the Carbon Footprint of BLOOM, a 176B Parameter Language Model. arXiv:2211.02001 [cs.LG] https://arxiv.org/abs/2211.02001

    2022 arXiv

  25. [33]

    McMillan

    James H. McMillan. 2007. Classroom assessment: Principles and practice that enhance student motivation and achievement. Journal of Scholarship of Teaching and Learning 7, 1 (2007), 22–33. https://files.eric.ed.gov/fulltext/EJ854925.pdf

    2007

  26. [34]

    Richard Nguyen. 2025. Bath Tub Icon. https://thenounproject.com/icon/bath- tub-7345415/. Accessed: 2025-01-23

    2025

  27. [35]

    James O’Donnell. 2025. DeepSeek might not be such good news for energy after all. https://www.technologyreview.com/2025/01/31/1110776/deepseek-might- not-be-such-good-news-for-energy-after-all/

    2025

  28. [36]

    Fety Rahmillah, Amina Tariq, Mark King, and Oscar Oviedo-Trespalacios. 2023. Evaluating the Effectiveness of Apps Designed to Reduce Mobile Phone Use and Prevent Maladaptive Mobile Phone Use: Multimethod Study. Journal of medical Internet research 25 (08 2023), e42541. https:/...

    2023 doi

  29. [37]

    Shaolei Ren and Adam Wierman. 2024. The uneven distribution of Ai’s envi- ronmental impacts. https://hbr.org/2024/07/the-uneven-distribution-of-ais- environmental-impacts

    2024

  30. [38]

    Donti, Lynn H

    David Rolnick, Priya L. Donti, Lynn H. Kaack, Kelly Kochanski, Alexandre Lacoste, Kris Sankaran, Andrew Slavin Ross, Nikola Milojevic-Dupont, Natasha Jaques, Anna Waldman-Brown, Alexandra Luccioni, Tegan Maharaj, Evan D. Sherwin, S. Karthik Mukkavilli, Konrad P. Kording, Carla...

    2019 arXiv

  31. [39]

    John Rooksby, Parvin Asadzadeh, Mattias Rost, Alistair Morrison, and Matthew Chalmers. 2016. Personal Tracking of Screen Time on Digital Devices. InProceed- ings of the 2016 CHI Conference on Human Factors in Computing Systems (San Jose, California, USA) (CHI ’16). Association...

    2016

  32. [40]

    Angela Sanguinetti, Kelsea Dombrovski, and Suhaila Sikand. 2018. Information, timing, and display: A design-behavior framework for improving the effectiveness of eco-feedback. Energy Research & Social Science 39 (2018), 55–68. https: //doi.org/10.1016/j.erss.2017.10.001

    2018 doi

  33. [41]

    Victor Schidt, Kamal Goyal, Aditya Joshi, Boris Feld, Liam Conell, Nikolas Laskaris, Doug Blank, Jonathan Wilson, Sorelle Friedler, and Sasha Luccioni. [n. d.]. CodeCarbon: estimate and track carbon emissions from machine learning computing. https://github.com/mlco2/codecarbon

  34. [42]

    Semianalysis. 2023. Peeling the Onion’s Layers: Large Language Models. https: //semianalysis.com/2023/02/13/peeling-the-onions-layers-large-language/ Ac- cessed: 2025-01-22

    2023

  35. [43]

    Shubham Singh. 2025. Number Of ChatGPT Users (January 2025). https: //www.demandsage.com/chatgpt-statistics/

    2025

  36. [44]

    Josh Staiger. 2024. Chatterclock — a ChatGPT message tracking extension. Chrome Extension Store (2024). https://chromewebstore.google.com/detail/ chatterclock-%E2%80%94-a-chatgpt/mepflplnjbngmgakdefimlgbfpmhonoj

    2024

  37. [45]

    Jaeger Stefano De Dominicis, Rebecca Sokoloski and P

    Christine M. Jaeger Stefano De Dominicis, Rebecca Sokoloski and P. Wesley Schultz. 2019. Making the Smart Meter Social Promotes Long-Term Energy Conservation. In Palgrave Commun, Vol. 5. https://doi.org/10.1057/s41599-019- 0254-5

    2019 doi

  38. [46]

    Jovan Stojkovic, Esha Choukse, Chaojie Zhang, Inigo Goiri, and Josep Torrellas

  39. [47]

    Emma Strubell, Ananya Ganesh, and Andrew McCallum. 2020. Energy and Policy Considerations for Modern Deep Learning Research. Proceedings of the AAAI Conference on Artificial Intelligence 34, 09 (Apr. 2020), 13693–13696. https: //doi.org/10.1609/aaai.v34i09.7123

    2020 doi

  40. [48]

    Brigitte Tousignant. 2021. This Climate Does Not Exist: Picturing impacts of the climate crisis with AI, one address at a time. Mila (October 2021)

    2021

  41. [49]

    arXiv:2403.20306 [cs.AI] https://arxiv.org/abs/2403.20306

    Towards Greener LLMs: Bringing Energy-Efficiency to the Forefront of LLM Inference. arXiv:2403.20306 [cs.AI] https://arxiv.org/abs/2403.20306

    arXiv

  42. [50]

    Johannes Tulusan, Lito Soi, Johannes Paefgen, Marc Brogle, and Thorsten Staake

  43. [51]

    Number of queries within study period: {study_query_count}

    Iconic Visions. 2025. Cybertruck Icon. https://thenounproject.com/icon/ cybertruck-7345261/. Accessed: 2025-01-23. GPTFootprint CHI EA ’25, April 26-May 1, 2025, Yokohama, Japan A TABULAR DATA Participant No. of Queries Per Week Before Trial No. of Queries Per Week During Tria...

    2025

  44. [52]

    Carbon Trust. 2025. SME Carbon Footprint Calculator. https: //www.carbontrust.com/our-work-and-impact/guides-reports-and-tools/sme- carbon-footprint-calculator Accessed: 2025-01-21

    2025

  45. [2007]

    JAMA : the journal of the American Medical Association 298 (11 2007), 2296–304

    Using Pedometers to Increase Physical Activity and Improve Health: A Systematic Review. JAMA : the journal of the American Medical Association 298 (11 2007), 2296–304. https://doi.org/10.1001/jama.298.19.2296

    2007 doi

  46. [2011]

    In2011 IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks

    Eco-efficient feedback technologies: Which eco-feedback types prefer drivers most?. In2011 IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks. 1–8. https://doi.org/10.1109/WoWMoM.2011.5986187

    2011

  47. [2024]

    arXiv:2405.21015 [cs.CY] https://arxiv.org/abs/2405.21015

    The rising costs of training frontier AI models. arXiv:2405.21015 [cs.CY] https://arxiv.org/abs/2405.21015

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.