Pith. sign in

REVIEW 4 major objections 5 minor 40 references

InvestESG: A multi-agent reinforcement learning benchmark for studying climate investment as a social dilemma

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read ESG disclosure alone does not drive corporate climate mitigation; a critical mass of ESG-conscious investors does.

desk verdict A genuinely reusable MARL benchmark for ESG disclosure with honest limitations, but its headline policy conclusions lean harder on a guessed climate-risk curve and thin seeds than the framing admits. read the letter →

arxiv 2411.09856 v3 pith:KFWQSAJH submitted 2024-11-15 cs.LG cs.CYcs.MAecon.GNq-fin.EC

Xiaoxuan Hou , Jiayi Yuan , Joel Z. Leibo , Natasha Jaques This is my paper · ORCID
classification cs.LGcs.CYcs.MAecon.GNq-fin.EC
keywords multi-agentreinforcementlearningESGdisclosureclimateinvestmentsocialdilemmagreenwashinginvestorpreferencesriskbenchmark
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

This paper builds a multi-agent reinforcement learning environment, InvestESG, in which companies choose how much to spend on mitigation, greenwashing, and resilience, while investors choose portfolios based on profit and ESG preferences. It claims that mandatory ESG disclosure alone leaves corporate mitigation limited when investors remain profit-driven. When enough investors with strong ESG preferences hold enough capital, companies learn to cooperate and reduce climate risk, improving long-term financial stability. The benchmark is offered to test policy and market designs through simulation rather than costly real-world experiments.

What carries the argument

The central object is a 100-year simulation environment with M companies allocating capital shares to mitigation, greenwashing, and resilience, and N investors choosing binary portfolios. The load-bearing mechanism is the climate risk transition $P_t^e = \frac{\mu_t^e}{1 + \lambda^e U_{t,m}} + P_0^e$, where $U_{t,m}$ is cumulative mitigation spending by all companies; with no mitigation, risk grows linearly toward the IPCC 4°C scenario, while sufficient spending bends it toward the 1.5°C scenario. Company ESG scores are $Q = u_m + \beta u_g$ with $\beta > 1$ making greenwashing cheaper per ESG point, and investor rewards add an ESG-weighted term scaled by the investor's preference $\alpha$. Schelling diagrams compare cooperating versus defecting payoffs as a function of the number of cooperating companies to diagnose when the environment is a social dilemma, and IPPO agents then learn policies from rewards.

What would settle it

The cleanest check is the paper's own status quo simulation: with investor ESG preference $\alpha = 0$ and the disclosure mandate on, the model predicts final climate risk essentially equal to the no-disclosure baseline of about 0.97. If a natural experiment, such as the EU's mandatory ESG reporting directive, shows large mitigation responses among firms with no measurable ESG-committed investor base, the central claim would be falsified.

Watch

Extended reading notes

Core claim

On the paper's own terms, climate change is an intertemporal social dilemma: companies pay the full short-term cost of mitigation but share the long-term benefit of reduced climate risk. Using Schelling diagrams and Independent PPO agents, the paper shows that a profit-driven investor base leaves the dilemma intact under an ESG disclosure mandate; only when a critical mass of investors with sufficiently high ESG-consciousness (α) and capital enters does mitigation become the individually rational choice, eliminating the dilemma in simulation. The paper further finds that providing agents with global climate-risk information raises mitigation even without investors, that greenwashing is initially explored by learning agents but largely abandoned when investors are slow to respond, and that resilience spending can support higher mitigation by keeping companies solvent.

Load-bearing premise

The functional form of Equation (1), linear risk growth with no mitigation damped multiplicatively by cumulative mitigation spending with coefficients fit to just two IPCC scenarios, is an unvalidated guess; if real mitigation returns differ, the trade-off that produces the critical-mass threshold is distorted and the policy conclusions could flip.

Editorial extensions

If this is right

  • Mandatory ESG disclosure should be paired with policies that strengthen the size and capital of the ESG-conscious investor base to be effective.
  • Market bifurcation is a likely equilibrium: a few mitigating companies attract ESG-focused capital while others free-ride for profit.
  • Providing companies with clear, system-wide climate risk information is a low-cost lever that increases mitigation even without investor pressure.
  • Greenwashing may be a smaller threat to disclosure effectiveness than feared, at least when investors are slow to adjust their strategies.
  • The benchmark offers a testbed for other policy variants, such as scope-specific disclosure, locked-in decisions, and stricter bankruptcy rules, before real-world enactment.

Reading between the lines

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

  • If the critical-mass effect holds, the policy implication is not just to mandate disclosure but to increase the capital share of ESG-committed institutional investors, for instance through fiduciary-duty clarification or public investment funds; the paper does not test such mechanisms.
  • The reward parameter $\alpha$ conflates preference with wealth; a model that varies investor capital concentration separately from $\alpha$ would test whether the 'sufficient capital' condition means number of investors or total assets under management.
  • The climate-damage functional form is the main sensitivity risk; a robustness suite varying $\lambda^e$ and the linear-growth baseline would reveal how much the critical-mass threshold depends on that assumption.
  • An immediate extension would make ESG-consciousness itself learnable by investors, as the paper lists for future work, and test whether self-regulation replaces the need for disclosure mandates.
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

4 major / 5 minor

Summary. The paper introduces InvestESG, a multi-agent reinforcement learning environment in which companies allocate capital among mitigation, greenwashing, and resilience over a simulated 100-year horizon, while investors choose portfolios that may reward ESG scores. The authors use Independent PPO and Schelling diagrams to argue that mandatory ESG disclosure alone does not induce mitigation when investors are purely profit-driven, that a critical mass of ESG-conscious investors does induce corporate cooperation and lower climate risk, that heterogeneous investor preferences produce market bifurcation, that additional climate-risk information increases mitigation even without investors, and that greenwashing does not significantly undermine learned mitigation behavior. The environment is released in both PyTorch and JAX.

Significance. If the results are robust, InvestESG is a useful benchmark for studying intertemporal social dilemmas and for comparing MARL algorithms on a policy-relevant problem. The paper's strengths include the open-source dual implementations, the use of Schelling diagrams to characterize the game structure, the scale-up experiments to 10 and 25 agents, and the alignment of several directional findings with empirical work on ESG disclosure. However, the headline result is substantially encoded in the reward design and in the ad hoc climate-risk equation, and the empirical evidence base is thin (three seeds, fixed climate-event seed, hand-set parameters). The paper is best read as a proof-of-concept benchmark rather than a validated policy model; with additional sensitivity analysis and statistical support it could become a solid contribution.

major comments (4)
  1. [Section 9, Eq. (1)] The climate-risk equation is the single most load-bearing modeling choice in the paper, because it defines the returns to mitigation that generate the social dilemma. The paper calibrates only two points: U=0 yields the 4C scenario by 2100, and a single annual $2.3T mitigation path yields the 1.5C scenario by 2100. The functional form P_t^e = mu^e t / (1 + lambda^e U_t,m) + P_0^e is then assumed to interpolate and extrapolate between and beyond these points, with no derivation, no alternative-form comparison, and no sensitivity analysis. If the actual or plausible relation between cumulative mitigation and risk reduction is convex, saturating, or threshold-like, the Schelling diagrams in Figure 3 and the learned policies in Figures 4-8 could change qualitatively, which would change the policy conclusions. Please add a robustness section that varies the functional form and recalibrates lambda^e, or at minimum states explicitly which conclusions are invariant to this choice.
  2. [Section 4.2 and Figures 4-8] All quantitative claims rest on three random seeds, and climate-event generation uses a fixed random seed across training episodes (footnote 5). With three seeds and no significance tests, the error bars in Figures 4, 6, 7, and 8 are not strong evidence for the central 'critical mass of ESG-conscious investors' claim, particularly for the small differences between Status Quo and Status Quo with Mandate. The fixed climate-event seed also means that climate-event realizations are not independently sampled across runs. Please report more seeds (or at least vary the climate-event seed), provide per-condition confidence intervals, and state whether observed differences are statistically reliable.
  3. [Section 5, Figures 6 and 9] The greenwashing result is internally inconsistent. The Schelling diagram in Figure 3e and the hard-coded experiment in Figure 9 show that greenwashing can attract ESG-conscious investors and re-create a social dilemma, yet the IPPO experiments in Figure 6 are used to conclude that greenwashing 'does not significantly undermine mitigation efforts.' The stated explanation is that investors learn too slowly to respond to greenwashing during early training, which means the result is an artifact of the co-adaptive training regime rather than a robust property of the environment. This should be tested explicitly, for example by pretraining investors to associate ESG scores with investment decisions, by increasing alpha, or by extending training, and the conclusion should be weakened to reflect the dependence on learning dynamics.
  4. [Section 9, Rewards and Eq. (3)] Part of the headline result is built into the reward function. The investor reward adds alpha times the weighted ESG score of the investor's portfolio, and the ESG score is, by Eq. (3), increasing in mitigation (and in greenwashing at rate beta). It is therefore true by construction that sufficiently high alpha favors mitigation, and the paper should state this explicitly rather than presenting 'more ESG-conscious investors increase mitigation' as an emergent empirical discovery. The novel content is the threshold behavior, the bifurcation with heterogeneous investors, and the interactions with greenwashing and resilience; the paper should focus the claims on those aspects.
minor comments (5)
  1. [Section 4.1, Figure 3] The text describing Figures 3d and 3e appears to be swapped relative to the captions: the paragraph discusses resilience spending for Figure 3e and greenwashing for Figure 3d, while the captions label (d) as resilience and (e) as greenwashing. Please correct the mismatch.
  2. [Section 9, Eq. (1)] Please define mu^e explicitly and give the exact calibration procedure for lambda^e; the current description ('the model fits lambda^e so that such investment levels would yield 1.5C scenario climate risks by 2100') is not reproducible without additional detail.
  3. [Throughout] There are several typos and grammatical errors, including 'a an intertemporal social dilemma' (Section 1), 'the environmental back into a social dilemma' (Section 4.1), 'acitions' (Appendix 11), 'strick bankruptcy mechanism' (Appendix 11.5), and 'long-term strateg' (Section 6). A careful proofread is needed.
  4. [Footnote 5] The justification for fixing the climate-event random seed is subjective ('this mirrors real-world baseline understanding'); please either provide evidence or rephrase this as a modeling convenience chosen to reduce learning variance.
  5. [Section 5, 'Providing additional information'] The information-provision result is reported only as final climate risk; please include learning curves and seed-level variation to show that the improvement is stable and not driven by a single run or by the fixed climate seed.

Circularity Check

1 steps flagged · score 6.0 of 10

Main ESG-investor result is hard-wired into reward definitions; greenwashing and information results are independent.

  1. self definitional [Section 4 (Rewards and Eq. 3); Section 5, 'Level of investors' ESG-consciousness increases mitigation efforts']
    "The reward for investor Ij is r^{Ij}_t = (K^{Ij}_{t+1}-K^{Ij}_t)/K^{Ij}_t + α^{Ij} (Σ_i H^{Ij}_{t+1,i} Q^{Ci}_{t+1})/(Σ_i K^{Ij}_{t+1}) ... the second component represents the weighted average ESG score of the investor’s portfolio adjusted by the investor’s ESG preference, α^{Ij} ... Q^{Ci}_{t+1} = u^{Ci}_{t,m} + β u^{Ci}_{t,g}, (3)."

    The paper's headline result—that raising investors' ESG-consciousness α increases corporate mitigation—follows directly from the definitions of the investor reward and the ESG score. An ESG-conscious investor is defined as one whose reward includes α times the portfolio's average ESG score, and Eq. (3) defines that ESG score as the company's mitigation spending (plus cheap greenwashing). Therefore any company that increases u_m (or u_g) mechanically increases the quantity that ESG-conscious investors are defined to maximize, giving a built-in incentive to mitigate. The Schelling diagrams in Fig. 3 and the α-ablation in Fig. 4 are computations of these same payoff equations, not emergent findings.

full rationale

InvestESG is an openly constructed simulation benchmark, and many of its results are sanity checks that the environment instantiates the intended incentive structure. The central abstract claim about ESG-conscious investors is hard-wired: investor utility is defined to reward ESG scores, and ESG scores are defined to reward mitigation. This is a self-definitional reduction of the main qualitative result. However, the paper also contains results not encoded in the reward functions—learned agents largely abandon greenwashing despite its cheap ESG benefit, and adding climate-risk information increases mitigation even with no investors—so the simulation is not equivalent to its inputs. There is no load-bearing self-citation chain or imported uniqueness theorem; citations to the authors' prior SSD work are background context. Score 6 reflects partial circularity in the main result rather than full equivalence.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The benchmark relies on several hand-set economic and climate parameters, most notably the mitigation efficiency lambda_e, which is calibrated to IPCC scenarios rather than derived from first principles. The behavioral axioms (profit-driven companies, exogenous ESG preferences) and the linear ESG-score formula are simplifying assumptions carried over from the cited economics literature. No new physical entities are introduced.

free parameters (5)
  • lambda_e (per-event mitigation efficiency) = calibrated so $2.3T annual mitigation yields IPCC 1.5C scenario by 2100
    Controls how strongly cumulative mitigation spending damps rising climate risk; fit to two IPCC scenarios, not derived from physical climate models.
  • beta (greenwashing coefficient) = 2, 10, 20 in experiments
    Defines how cheaply greenwashing inflates ESG scores; varied by hand to test sensitivity, no independent empirical grounding.
  • alpha (investor ESG consciousness) = 0, 0.5, 1, 10
    Weights ESG utility in investor reward; chosen by hand to represent different investor preferences, not fitted.
  • eta (resilience diminishing return rate) = not specified in main text
    Sets how quickly resilience investment becomes less effective; value not given, likely hand-set.
  • gamma (baseline growth rate) = 10%
    Aligned with S&P 500 historical return; exogenous input rather than fitted to the simulation.
assumptions (4)
  • ad hoc to paper Climate risk grows linearly without mitigation and is damped by cumulative mitigation via Eq. (1).
    The functional form is chosen by the authors and calibrated to two IPCC scenarios; it is not derived from a physical climate model.
  • domain assumption Company agents are purely profit-driven and maximize one-period profit margins.
    Basis of the social dilemma; companies do not internalize climate externalities except through profit and investor reactions.
  • domain assumption Investor rewards are a linear combination of portfolio return and the portfolio's weighted average ESG score.
    Adapted from Pastor et al. (2021) and Pedersen et al. (2021); assumes ESG preferences are exogenous and fixed.
  • ad hoc to paper ESG scores are a linear function of mitigation and greenwashing spending (Eq. 3).
    Simplified representation of ESG ratings; no evidence that real ESG scores follow this exact linear form.

how reviews work

0 comments
Cite this review

Pith. "Pith review of InvestESG: A multi-agent reinforcement learning benchmark for studying climate investment as a social dilemma." pith.science (2026). https://pith.science/paper/KFWQSAJH

@misc{pith2026241109856,
  author       = {Pith},
  title        = {Pith review of: InvestESG: A multi-agent reinforcement learning benchmark for studying climate investment as a social dilemma},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KFWQSAJH}},
  note         = {Machine review of arXiv:2411.09856}
}
read the original abstract

InvestESG is a novel multi-agent reinforcement learning (MARL) benchmark designed to study the impact of Environmental, Social, and Governance (ESG) disclosure mandates on corporate climate investments. The benchmark models an intertemporal social dilemma where companies balance short-term profit losses from climate mitigation efforts and long-term benefits from reducing climate risk, while ESG-conscious investors attempt to influence corporate behavior through their investment decisions. Companies allocate capital across mitigation, greenwashing, and resilience, with varying strategies influencing climate outcomes and investor preferences. We are releasing open-source versions of InvestESG in both PyTorch and JAX, which enable scalable and hardware-accelerated simulations for investigating competing incentives in mitigate climate change. Our experiments show that without ESG-conscious investors with sufficient capital, corporate mitigation efforts remain limited under the disclosure mandate. However, when a critical mass of investors prioritizes ESG, corporate cooperation increases, which in turn reduces climate risks and enhances long-term financial stability. Additionally, providing more information about global climate risks encourages companies to invest more in mitigation, even without investor involvement. Our findings align with empirical research using real-world data, highlighting MARL's potential to inform policy by providing insights into large-scale socio-economic challenges through efficient testing of alternative policy and market designs.

Figures

Figures reproduced from arXiv: 2411.09856 by the authors.

Figure 1
Figure 1. The InvestESG Environment. Corporations choose how much to invest in mitigating emissions, which affects their ESG Score. Climate-conscious investors can see ESG Scores when deciding how much to invest in each company. However, companies can engage in greenwashing to inexpensively and falsely improve ESG scores without actually mitigating climate change. InvestESG is a social dilemma, where selfish, profit￾motivated… view at source ↗
Figure 2
Figure 2. Status quo scenario where all agents are only profit-motivated. In (a), mitigation spending (blue curve) is minimal, leading climate risk (yel￾low curve) to increase over time. Ad￾verse weather event occurrences are shown as dotted lines; red lines indi￾cate multiple adverse events in a sin￾gle year. (b) plots the average number severe climate events over the episode in (a), showing how increasing climate risk leads… view at source ↗
Figure 3
Figure 3. Schelling diagrams demonstrating that the environment constitutes a social dilemma. The graphs compare payoffs between cooperation (mitigation, blue lines) and defection (no mitigation, red lines) for a focal company, given varying number of other cooperating companies. Yellow lines represent the average payoff across all companies when the focal company defects. Subfigure (a) illustrates the selfish scenario, where… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Ending values for all metrics averaged over the last 100 episodes; error bars show std. err. over 3 random seeds. We compare the status quo scenario with solely profit-driven investors (investors with ESG consciousness level of 0), both with and without the ESG disclos…
Figure 5
Figure 5. Figure 5: Investigating the effects of the level of ESG consciousness in the case of 5 companies and 3 investors, where investor 0 is profit driven (α I0 = 0), and investors 1 and 2 are deeply climate-conscious (α I1 = α I2 = 10). In (a) and (b), Company 0 (purple) learns to be …
Figure 6
Figure 6. Figure 6: Mitigation, greenwashing, and climate risk for greenwashing coefficient = 2, 10, 20, with initial exploration episodes zoomed in in (b). It shows that when both companies and investors are IPPO agents, regardless of greenwashing cost, companies initially explore greenw…
Figure 7
Figure 7. Figure 7: Effect of providing more information about climate risk to both investors and companies in the default 5-company-3-investor case (a), or companies only in a 5-company-0-investor case (b). These results show that simply providing more information about climate risk to c…
Figure 8
Figure 8. Figure 8: Effect of allowing resilience spending. When resilience is allowed, companies invest significantly more in resilience than in mitigation, resulting in ending episode higher market wealth, and lower climate risk. calculation of ESG scores. As shown in [PITH_FULL_IMAGE:…
Figure 9
Figure 9. Figure 9: ESG-conscious investors can be distracted by greenwashing, and heavily invest in a company that greenwashes. Here we examine a scenario where Company 0 is hard-coded to invest in real mitigation while Company 1 only invests in greenwashing. Investors have ESG conscious…
Figure 10
Figure 10. Figure 10: The learning curve for climate risk, total market wealth, total mitigation amount and climate event occurrence over the course of training for the 5-company-3-investor case. We compare the status quo scenario with solely profit-driven investors (investors with ESG con…
Figure 11
Figure 11. Figure 11: (a)(b) shows the final climate risk and market total wealth for the case of 25 companies and 25 investors. Similar to the default 5-company-and-3-investor case, when investors are highly conscious, the final climate risk would be decreased. (c)(d) shows the ending epi…
Figure 12
Figure 12. Figure 12: (a)-(c) Effect of seeding with real-world data. (d)-(f) Effect of 5-year lock-in period for agent de￾cisions. All comparisons are made against the default case with an ESG disclosure mandate, in which investors have zero ESG-consciousness, and company acitions are res…
Figure 13
Figure 13. Figure 13: (a)-(c) Effect of uncertain climate event damage. (d)-(g) Effect of uncertain climate event dam￾age on greenwash spending. (h)-(k) Effect of a more strick bankruptcy mechanism and its combination with uncertain climate event damage. (a) ESG consciousness = 0.5 (b) ESG…
Figure 14
Figure 14. Figure 14: (a)-(c) shows the ending episode investment matrix for investors with different level of ESG consciousness. (d) shows ending climate risk vs. number of companies under the setting where no investors are ESG-conscious. When the number of companies increases, each compa…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 28 canonical work pages

  1. [2]

    Controlling behavioral diversity in multi- agent reinforcement learning

    Matteo Bettini, Ryan Kortvelesy, and Amanda Prorok. Controlling behavioral diversity in multi- agent reinforcement learning. arXiv preprint arXiv:2405.15054, 2024a. Matteo Bettini, Amanda Prorok, and Vincent Moens. Benchmarl: Benchmarking multi-agent rein- forcement learning. Journal of Machine Learning Research, 25(217):1–10, 2024b. Alexander Bisaro and ...

    arXiv

  2. [4]

    Accessed 27-11-2024

    URL https://www.climatepolicyini tiative.org/press-release/annual-finance-for-climate-action-sur passes-usd-1-trillion-but-far-from-levels-needed-to-avoid-devas tating-future-losses/. Accessed 27-11-2024. Micah Carroll, Rohin Shah, Mark K. Ho, Thomas L. Griffiths, Sanjit A. Seshia, Pieter Abbeel, and Anca Dragan. On the utility of learning about humans fo...

    work page 2024

  3. [6]

    Denis Cormier and Michel Magnan

    URL https://www.cnbc.com/2024/03/04/the-sec-votes-this-week-on-c ontroversial-climate-change-rule-heres-whats-at-stake.html . Denis Cormier and Michel Magnan. Corporate environmental disclosure strategies: determinants, costs and benefits. Journal of Accounting, Auditing & Finance, 14(4):429–451,

    work page 2024

  4. [7]

    Is independent learning all you need in the starcraft multi-agent challenge? arXiv preprint arXiv:2011.09533,

    Christian Schroeder De Witt, Tarun Gupta, Denys Makoviichuk, Viktor Makoviychuk, Philip HS Torr, Mingfei Sun, and Shimon Whiteson. Is independent learning all you need in the starcraft multi-agent challenge? arXiv preprint arXiv:2011.09533,

    arXiv 2011

  5. [10]

    11 A DDITIONAL EXPERIMENTS 11.1 S CALE UP THE NUMBER OF AGENTS Figure 11a - 11b show the experiment results when the number of company and investor agents are scaled up to 25-by-25. The increased number of agents reveal the same directional story as the main results shown in Figure 4, where highly ESG-conscious investors motivate mitigation efforts from c...

    work page 2023

  6. [11]

    Accessed: 2024-11-26

    URL https://www.eib.org/en/publications/20230114-what-drives-firms-i nvestment-in-climate-change . Accessed: 2024-11-26. Eugene F Fama and Kenneth R French. Disagreement, tastes, and asset prices. Journal of financial economics, 83(3):667–689,

    work page arXiv 2024

  7. [13]

    Paul Griffin and CR Heede

    Accessed: 2024-11-26. Paul Griffin and CR Heede. The carbon majors database. CDP carbon majors report 2017 , 14,

    work page 2024

  8. [15]

    Korea to enhance esg disclosures as part of corporate sustainability efforts,

    13 Published as a conference paper at ICLR 2025 Korea Economic Daily. Korea to enhance esg disclosures as part of corporate sustainability efforts,

    work page 2025

Show all 40 references
  1. [16]

    Accessed: 2024-08-22

    URL https://www.kedglobal.com/esg/newsView/ked202310160022 . Accessed: 2024-08-22. Philipp Krueger, Zacharias Sautner, Dragon Yongjun Tang, and Rui Zhong. The effects of mandatory esg disclosure around the world. Journal of Accounting Research,

    2024

  2. [17]

    Multi-agent reinforcement learning in sequential social dilemmas

    Joel Z Leibo, Vinicius Zambaldi, Marc Lanctot, Janusz Marecki, and Thore Graepel. Multi-agent reinforcement learning in sequential social dilemmas. arXiv preprint arXiv:1702.03037,

    arXiv

  3. [19]

    Christopher Marquis, Michael W Toffel, and Yanhua Zhou

    URL https://arxiv.org/abs/2103.03104. Christopher Marquis, Michael W Toffel, and Yanhua Zhou. Scrutiny, norms, and selective disclo- sure: A global study of greenwashing. Organization Science, 27(2):483–504,

    arXiv

  4. [20]

    Masson-Delmotte, P

    V . Masson-Delmotte, P. Zhai, A. Pirani, S.L. Connors, C. P ´ean, S. Berger, N. Caud, Y . Chen, L. Goldfarb, M.I. Gomis, M. Huang, K. Leitzell, E. Lonnoy, J.B.R. Matthews, T.K. Maycock, T. Waterfield, O. Yelekc ¸i, R. Yu, and B. Zhou (eds.).Climate Change 2021: The Physical Sc...

    2021

  5. [21]

    Accessed: 2024-09-19

    URL https://www.msci.com/sustainable-inves ting/esg-ratings. Accessed: 2024-09-19. William D Nordhaus. An optimal transition path for controlling greenhouse gases. Science, 258 (5086):1315–1319,

    2024

  6. [22]

    climateimpact.com/news-insights/fortune-global-500-climate-commi tments/

    URL https://www. climateimpact.com/news-insights/fortune-global-500-climate-commi tments/. Accessed: 2024-11-26. L’uboˇs P ´astor, Robert F Stambaugh, and Lucian A Taylor. Sustainable investing in equilibrium. Journal of financial economics, 142(2):550–571,

    2024

  7. [23]

    Automatic differentiation in pytorch

    14 Published as a conference paper at ICLR 2025 Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan, Edward Yang, Zachary DeVito, Zeming Lin, Alban Desmaison, Luca Antiga, and Adam Lerer. Automatic differentiation in pytorch. In NIPS-W,

    2025

  8. [26]

    Mikayel Samvelyan, Tabish Rashid, Christian Schroeder De Witt, Gregory Farquhar, Nantas Nardelli, Tim GJ Rudner, Chia-Man Hung, Philip HS Torr, Jakob Foerster, and Shimon Whiteson

    URL http://jmlr.org/papers/v22/20-136 4.html. Mikayel Samvelyan, Tabish Rashid, Christian Schroeder De Witt, Gregory Farquhar, Nantas Nardelli, Tim GJ Rudner, Chia-Man Hung, Philip HS Torr, Jakob Foerster, and Shimon Whiteson. The starcraft multi-agent challenge. arXiv preprin...

    1902 arXiv

  9. [27]

    The enhancement and standardization of climate-related disclosures for investors; delay of effective dates, 2024a

    SEC. The enhancement and standardization of climate-related disclosures for investors; delay of effective dates, 2024a. URL https://www.federalregister.gov/documents/20 24/04/12/2024-07648/the-enhancement\protect\penalty\z@-and-stand ardization-of-climate-related-disclosures-f...

    2024

  10. [28]

    doi: 10.3926/ic.1354

    ISSN 1697-9818. doi: 10.3926/ic.1354. URL https://www.intangiblecapital.org/index.php/ic/a rticle/view/1354. P.R. Shukla, J. Skea, R. Slade, A. Al Khourdajie, R. van Diemen, D. McCollum, M. Pathak, S. Some, P. Vyas, R. Fradera, M. Belkacemi, A. Hasija, G. Lisboa, S. Luz, and J...

    doi

  11. [29]

    more than words

    doi: 10.1017/9781009157926. Alfonso Siano, Agostino V ollero, Francesca Conte, and Sara Amabile. “more than words”: Expand- ing the taxonomy of greenwashing after the volkswagen scandal. Journal of business research, 71:27–37,

    doi

  12. [30]

    Accessed: 2024-09-19

    URL https://www.marketplace.spglobal .com/en/datasets/s-p-global-esg-scores-(171) . Accessed: 2024-09-19. WFE Statistics. Market Capitalisation Q3

    2024

  13. [31]

    Accessed: 2024-09-27

    URL https://focus.world-exchanges.org/articles/market-capitalisatio n-q3-2023. Accessed: 2024-09-27. Yue Wu, Kaifu Zhang, and Jinhong Xie. Bad greenwashing, good greenwashing: Corporate social responsibility and information transparency. Management Science, 66(7):3095–3112,

    2023

  14. [32]

    Learning to incentivize other learning agents

    Jiachen Yang, Ang Li, Mehrdad Farajtabar, Peter Sunehag, Edward Hughes, and Hongyuan Zha. Learning to incentivize other learning agents. CoRR, abs/2006.06051, 2020a. URL https: //arxiv.org/abs/2006.06051. Zhi Yang, Thi Thu Huong Nguyen, Hoang Nam Nguyen, Thi Thuy Nga Nguyen, a...

    2006 arXiv

  15. [33]

    Ai for global climate cooperation: modeling global climate negotiations, agreements, and long-term cooperation in rice-n

    Tianyu Zhang, Andrew Williams, Soham Phade, Sunil Srinivasa, Yang Zhang, Prateek Gupta, Yoshua Bengio, and Stephan Zheng. Ai for global climate cooperation: modeling global climate negotiations, agreements, and long-term cooperation in rice-n. arXiv preprint arXiv:2208.07004,

    arXiv

  16. [34]

    16 Published as a conference paper at ICLR 2025 In the Appendix, we will cover the following details of our work

    URL https://arxiv.org/abs/2108.02755. 16 Published as a conference paper at ICLR 2025 In the Appendix, we will cover the following details of our work. • Technical details of the InvestESG environment Appendix 9: Math formulation of the environment. • Implementation details Ap...

    2025 arXiv

  17. [36]

    Xt = Xt,h + Xt,p + Xt,e, where Xt,e ∼ Bernoulli(P e t ), for e ∈ {h, p, d}. (2) In addition to the evolving climate risks, the environment incorporates a baseline economic growth rate γ, set to 10% by default, aligned with the historical average annual return of the S&P 500 ov...

    2025

  18. [37]

    State and Observation Space

    or use letter-based ratings (MSCI Inc., 2024). State and Observation Space. The environment simulates a partially observable Markov game M defined over a continuous, multi-dimensional state space. The system state at period t is character- ized by the three climate risk parame...

    2024

  19. [38]

    Ut,m = Ut−1,m + MX i=1 uCi t,m × K Ci t+1,interim (5) U Ci t,r = U Ci t−1,r + uCi t,r × K Ci t+1,interim for i = 1,

    Here, Ut,m represents the cumulative mitigation spending by all company agents up to period t, while U Ci t,r denotes the cumulative resilience spending by company Ci. Ut,m = Ut−1,m + MX i=1 uCi t,m × K Ci t+1,interim (5) U Ci t,r = U Ci t−1,r + uCi t,r × K Ci t+1,interim for ...

    2025

  20. [39]

    Each agent has its own policy parameters, and agents do not share parameters among themselves

    10 I MPLEMENTATION DETAILS 10.1 I NDEPENDENT -PPO To test how self-interest agents learn to respond to incentives in the environment, we employ a state- of-the-art MARL algorithm based on Independent PPO. Each agent has its own policy parameters, and agents do not share parame...

  21. [41]

    To reflect the capital inflexibility, we implemented a 5-year lock-in period for agent decisions

    11.3 L OCK -IN INVESTMENTS In reality, the decision-making processes of both companies and investors can be less flexible than modeled, where companies and investors update their strategies annually. To reflect the capital inflexibility, we implemented a 5-year lock-in period ...

    2016

  22. [100]

    Company agents can mitigate the growth of climate risk by investing in emissions reduction

    Figure 2a depicts how increased climate risks and adverse climate events increase over time in a scenario where companies are solely profit-motivated. Company agents can mitigate the growth of climate risk by investing in emissions reduction. The change in climate risk P e t f...

    2022

  23. [2011]

    Nearly half of fortune 500 companies engaged in major climate initiatives

    David Gardiner and Associates. Nearly half of fortune 500 companies engaged in major climate initiatives. https://www.dgardiner.com/fortune-500-climate-initiativ es-2023/#:˜:text=There%20are%20now%20239%20Fortune,of%20the%20U. S.%20Fortune%20500,

    2023

  24. [2012]

    P ¨ortner, D.C

    H.-O. P ¨ortner, D.C. Roberts, M. Tignor, E.S. Poloczanska, K. Mintenbeck, A. Alegr ´ıa, M. Craig, S. Langsdorf, S. L ¨oschke, V . M¨oller, A. Okem, and B. Rama (eds.). Climate Change 2022: Im- pacts, Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth...

    2022

  25. [2013]

    Durugkar, Clemens Rosenbaum, Stefan Dernbach, and Sridhar Mahadevan

    12 Published as a conference paper at ICLR 2025 Ishan P. Durugkar, Clemens Rosenbaum, Stefan Dernbach, and Sridhar Mahadevan. Deep reinforce- ment learning with macro-actions,

    2025

  26. [2016]

    Javier El-Hage

    URL https://arxiv.org/abs/1606.04615. Javier El-Hage. Fixing esg: Are mandatory esg disclosures the solution to misleading esg ratings? Fordham J. Corp. & Fin. L., 26:359,

    arXiv

  27. [2019]

    Ortega, DJ Strouse, Joel Z

    Natasha Jaques, Angeliki Lazaridou, Edward Hughes, C ¸ aglar G ¨ulc ¸ehre, Pedro A. Ortega, DJ Strouse, Joel Z. Leibo, and Nando de Freitas. Intrinsic social motivation via causal influ- ence in multi-agent RL. CoRR, abs/1810.08647,

    arXiv

  28. [2020]

    Yi-Chun Chen, Mingyi Hung, and Yongxiang Wang

    URL https://arxiv.org/abs/1910.05789. Yi-Chun Chen, Mingyi Hung, and Yongxiang Wang. The effect of mandatory csr disclosure on firm profitability and social externalities: Evidence from china. Journal of accounting and economics, 65(1):169–190,

    1910 arXiv

  29. [2021]

    What drives firms’ investment in climate action? evidence from the 2022-2023 eib investment survey

    European Investment Bank. What drives firms’ investment in climate action? evidence from the 2022-2023 eib investment survey. Technical report, European Investment Bank,

    2022

  30. [2022]

    Antonin Raffin, Ashley Hill, Adam Gleave, Anssi Kanervisto, Maximilian Ernestus, and Noah Dor- mann

    doi: 10.1017/9781009325844. Antonin Raffin, Ashley Hill, Adam Gleave, Anssi Kanervisto, Maximilian Ernestus, and Noah Dor- mann. Stable-baselines3: Reliable reinforcement learning implementations. Journal of Machine Learning Research, 22(268):1–8,

    doi

  31. [2023]

    Amir Amel-Zadeh and George Serafeim

    URL https://arxiv.org/abs/2211.13746. Amir Amel-Zadeh and George Serafeim. Why and how investors use esg information: Evidence from a global survey. Financial analysts journal, 74(3):87–103,

    arXiv

  32. [2024]

    11 Published as a conference paper at ICLR 2025 Frances E Bowen

    URL https://www.bloomberg.com/news/articles/2024-02-07/eu-to-delay-e sg-reporting-rule-for-some-sectors-by-two-years?embedded-check out=true. 11 Published as a conference paper at ICLR 2025 Frances E Bowen. Environmental visibility: a trigger of green organizational response? ...

    2024

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.