Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Autonomous Alignment with Human Value on Altruism through Considerate Self-imagination and Theory of Mind

T0 review · 5 major / 6 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read An RL agent can learn to rescue others and avoid side effects from imagination alone.

desk verdict A clean proof-of-concept for combining AUP-style side-effect penalties with a random-reward ToM empathy term, but the 'autonomous alignment' framing outruns the evidence and the empathy proxy needs sharper justification. read the letter →

arxiv 2501.00320 v2 pith:VHP62NS2 submitted 2024-12-31 cs.AI

Haibo Tong , Enmeng Lu , Yinqian Sun , Zhengqiang Han , Chao Liu , Feifei Zhao , Yi Zeng This is my paper · ORCID
classification cs.AI
keywords altruismTheoryofMindself-imaginationnegativesideeffectsintrinsicmotivationreinforcementlearningAIsafetyvaluealignment
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 claims that a reinforcement-learning agent can align with human altruistic values without being given explicit rewards for altruism or for avoiding side effects. The key idea is to give the agent a self-imagination module: a set of value functions trained with random rewards on the agent's own experience, which let it predict the consequences of its actions on the environment and on another agent. Using these imagined values, the agent computes two intrinsic rewards: one that penalizes actions that lower the value of the environment, and one that rewards actions that raise the value of another agent's state. In a grid-world inspired by a story of breaking a vat to save a trapped child, the agent learns to prioritize rescuing the trapped person, to avoid smashing extra vats, and still reach its goal, while two baseline methods fail at one or more of these tasks. If correct, this shows a general mechanism for value alignment that is task-independent and does not require hand-specified auxiliary rewards.

What carries the argument

The key object is the self-imagination module: $N$ value functions $Q_i$, each trained as if the environment gave a uniformly random reward in $[0,1)$, updated from the agent's real interactions. The stepwise inaction baseline $\emptyset$ is the reference point: an action's effect is measured by $Q_i(s,a) - Q_i(s,\emptyset)$. The negative-side-effect penalty $R_{\mathrm{nse}}$ averages only the negative parts of these differences across the $N$ value functions, and the empathy incentive $R_{\mathrm{emp}}$ averages the same differences evaluated at the other agent's state $s_{\mathrm{other}}$, which is how the paper implements Theory of Mind: the agent uses its own imagined values as a proxy for the other's well-being. The total reward integrates these with the environment reward, and the whole loop--real experience updates the $Q_i$, which shape the intrinsic rewards, which guide the policy--runs online.

What would settle it

Construct an environment where the other agent's true utility is known--for example, the other agent actually wants to remain trapped--and check whether the sign of $R_{\mathrm{emp}}(s,a)$ matches the sign of the true utility difference for the other agent in each state; a systematic mismatch would falsify the claim that the agent's own $Q_i$ captures the other's welfare.

Watch

Extended reading notes

Core claim

The central discovery is that a single set of random-reward Q-functions, learned on the agent's own transitions, can serve two purposes at once: quantifying the negative side effect of an action (by comparing $Q_i(s,a)$ with $Q_i(s,\emptyset)$) and quantifying the effect on another agent (by applying the same $Q_i$ to the other's state). The paper shows that combining the resulting penalty term $R_{\mathrm{nse}}$ and empathy term $R_{\mathrm{emp}}$ with the environment reward, via $R_{\mathrm{total}} = \frac{R_{\mathrm{env}} - \alpha R_{\mathrm{nse}} + \beta R_{\mathrm{emp}}}{(\alpha+\beta)/2}$, produces a policy that, in all six tested environments, rescues a trapped human when one exists, avoids smashing vats when no one is trapped, and reaches the goal; in the most complex environment the agent even takes a detour to smash only the vat containing the human, avoiding an unnecessary adjacent vat. The ablation experiments show that removing either intrinsic term breaks the balance, and the method is stable across hyperparameter values and works with spiking neural networks, which the paper takes as evidence of generalizability.

Load-bearing premise

The load-bearing assumption is that the value functions an agent learns for its own experience under random rewards faithfully measure how much an action helps or harms another agent; if the other agent's interests fall outside what those random rewards capture, the empathy reward could reward actions that are not actually altruistic.

Editorial extensions

If this is right

  • An agent trained with only a goal reward and a time penalty will, under this framework, spontaneously prioritize rescuing another agent over its own goal, and will avoid unnecessary damage to the environment.
  • Because the self-imagination module is built from random rewards, it carries no assumptions about the task or the environment, so the same intrinsic reward mechanism should transfer to other tasks and domains without retuning.
  • The framework is architecture-agnostic: replacing the deep network with a spiking neural network preserves the behavior, indicating the intrinsic motivation terms can be layered onto existing reinforcement-learning algorithms.
  • Ablation results suggest that the empathy term alone can drive both rescuing and side-effect avoidance in some environments, because preserving the environment also preserves another's interests; the side-effect term is most needed when no other agent is present.

Reading between the lines

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

  • A natural stress test is to vary the other agent's actual goal: if the other agent's welfare is not aligned with what the random-reward Q-functions reward, the empathy term could reward harmful actions; the paper's grid-world only includes a trapped human whose welfare aligns with being freed.
  • The stepwise inaction baseline treats doing nothing as free, but in some environments inaction is itself irreversible or costly; the framework may need modification there.
  • The paper's Theory of Mind is a single-agent proxy--it uses the agent's own $Q_i$ to evaluate the other's state; a richer ToM that models the other's actual reward function, beliefs, or policy could extend this to competitive or deceptive settings.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes a framework called Considerate Self-imagination and Theory of Mind for autonomous alignment with altruistic values. The method trains multiple Q-functions on random reward functions in an imagined space, then derives two intrinsic rewards: Rnse, which penalizes actions that reduce imagined state values relative to an inaction baseline, and Remp, which rewards actions that increase the imagined value of the other agent's state relative to inaction. These intrinsic rewards are combined with the environmental reward in Eq. (4). The approach is evaluated in a grid-world 'Smash the Vat' environment and five variants that create conflicts among reaching a goal, avoiding smashing vats, and rescuing a trapped other agent. Experiments report that the combined method outperforms DQN and Empathy DQN on all three objectives, with ablations showing the individual contributions of Rnse and Remp, and additional experiments on hyperparameters and SNN compatibility.

Significance. If the central claim holds, the paper offers a task-independent intrinsic motivation mechanism that balances self-goal achievement, other-agent welfare, and side-effect avoidance without explicit rewards for the latter two, which would be a valuable contribution to AI safety and alignment. The authors provide a reproducible environment and reference implementation, and the ablation study clearly demonstrates that both Rnse and Remp contribute to the observed behavior. The main empirical result—that the full reward achieves all three objectives in the tested grid-worlds—is supported by the reported trajectories and quantitative curves. However, the paper's broader claim of 'autonomous alignment with human altruistic values' is substantially stronger than what is demonstrated, and the theoretical grounding of the empathy proxy is thin. The work is best viewed as a preliminary proof-of-concept in a specific, small environment.

major comments (5)
  1. [Section 4.3, Eq. (3)] The empathy term Remp(s,a) in Eq. (3) does not measure the effect of the agent's action on the other agent's welfare: Qi is trained on the agent's own MDP, and Qi(s_others,a) evaluates the value of the agent hypothetically being in the other's state and taking action a, not the causal value to the other of the agent's action. Because s_others^{t+1} never appears in Eq. (3), the term cannot capture transition-induced gains or losses for the other agent, and because random rewards are independent of the other's goals, high Qi values need not correspond to the other's well-being. The assertion in Section 4.3 that 'it is reasonable to directly use the same Qi to estimate the value of others' state' is not derived or tested beyond this grid-world, and the central claim of autonomous altruistic alignment rests on this proxy.
  2. [Section 4.1, Eq. (1)] The update rule in Eq. (1) is not standard Q-learning: Qi(st,at) is replaced by the maximum of its current value and the Bellman target, so Qi is monotonically nondecreasing and has no convergence guarantee. Since Rnse and Remp in Eqs. (2)-(3) are linear in Qi, unbounded or oscillating Qi would make the intrinsic rewards unstable, and the paper provides no boundedness or convergence analysis. This is a load-bearing technical gap for the proposed intrinsic reward computation.
  3. [Section 4.4, Eq. (4)] The normalization in Eq. (4) changes the scale of the environmental reward Renv whenever α or β change: for α=β=1 the denominator is 1, while for α=β=20 the denominator is 20, so the effective weight of Renv is divided by 20. Consequently, the hyperparameter experiments in Section 2.2.4 confound the relative weights of the three reward components with the overall scale of Renv, and the reported robustness may be an artifact of this scaling rather than of the method's insensitivity.
  4. [Table 3] In CShapeVatGoalEnv, the full Rtotal achieves a vat remain rate of only 0.666, which is well below the near-perfect rates in BasicVatGoalEnv (0.992) and matches the Renv+Rnse ablation (0.666). This indicates that in an environment without a trapped human, the full method does not reliably avoid smashing vats, weakening the claim that the agent 'avoids the negative effects of smashing the vat as a secondary target' across all six environments.
  5. [Section 3] The Discussion states that 'the proposed intrinsic incentive mechanism is not predicated on enabling agents to comprehend human morality' and that the environment 'remains insufficiently complex,' which directly qualifies the abstract's claim that the framework 'enables agents to autonomously align with human altruistic values.' The central claim should be re-scoped to alignment with a hand-specified altruistic objective in simple grid-worlds, or additional evidence must be provided for transfer to more complex settings.
minor comments (6)
  1. [Section 2.2.3] In the ablation description, 'When considering only the the empathy incentive term Rnse' uses the wrong symbol (Rnse should be Remp), and a few lines later 'the empathy incentive term Rnse' repeats the same typo.
  2. [Section 2.2.5] The sentence 'our performs well when integrated with SNN' is missing a noun; it should read 'our method performs well...'
  3. [Section 4.3] The sentence 'we define the inherent empathy incentive term Rnse(s, a)' should refer to Remp(s,a) to match Eq. (3).
  4. [General] The manuscript references Fig. 2 with trajectories for each environment, but the figure content is not reproduced in the text provided; please ensure Figure 2 is included in the final submission.
  5. [Section 2.2.4] The hyperparameter analysis states that two scenarios with unequal α and β were tested, but the specific values and results are not described; adding this information would improve reproducibility.
  6. [Eq. (4)] The symbols α and β are introduced without recommended ranges or a sensitivity discussion beyond the equal-value cases; a brief guideline would help readers apply the method.

Circularity Check

1 steps flagged · score 4.0 of 10

The empathy incentive (Eq. 3) is inserted directly into the optimized reward (Eq. 4), so the headline altruistic-rescue result is the construction's own objective; the random-reward Qi learning and external DQN/Empathy-DQN comparisons remain genuinely independent, making the circularity partial.

  1. self definitional [Section 4.3 (Eq. 3), Section 4.4 (Eq. 4), Section 2.2.1 (results), Section 3 (limitation)]
    "we define the inherent empathy incentive term ... Remp(s, a) := 1/N Σ (Qi(sothers, a) − Qi(sothers, ∅)) (3) ... Rtotal(s, a) := (Renv(s, a) − αRnse(s, a) + βRemp(s, a)) / ((α + β)/2) (4) ... it is reasonable to directly use the same Qi to estimate the value of others' state Qi(sothers, a), which reflects the essence of empathy."

    Eq. 4 adds +βRemp to the reward the policy maximizes, and Eq. 3 defines Remp as the improvement in the other-agent state under the agent's own Qi. Optimizing Rtotal thus maximizes the paper's operational measure of empathy: the reported outcome ('the agent prioritizes rescuing people by smashing the vat,' Section 2.2.1) is the injected reward's own objective; the ablation confirms the reduction, since rescue rate reaches ~1.0 exactly when Remp is present (Table 3).

full rationale

The derivation chain: Qi (Eq. 1) is learned from random rewards on the agent's own real-environment transitions; this component is genuinely independent of the target behavior and is adopted from the external AUP line ([18], [19]). The hand-designed steps are Eqs. 2-4: Rnse penalizes Q-drops (side effects) and Remp rewards Q-gains evaluated at the other agent's state (Eq. 3); both are injected into the optimized Rtotal (Eq. 4) with tunable weights. The main experimental claim, that the agent 'prioritizes rescuing people by smashing the vat' (Section 2.2.1), is produced by Remp: the ablation (Table 3) shows human-rescue rate rising to about 1.0 only when Remp is present. Because Remp is defined as the quantity whose maximization is then reported as empathy, the observation 'the agent improves the other's state value' is the reward term doing its job; the paper's vocabulary ('autonomously align,' 'spontaneous,' 'implicit intrinsic motivations,' 'essence of empathy') frames this injected objective as an emergent ToM capacity. The circularity is only partial: whether Qi(s_others, a) − Qi(s_others, ∅) happens to favor the rescue action is an empirical fact about learned Q-functions (random rewards are independent of the target), and the proxy demonstrably diverges from ideal altruism (SmashAndDetourEnv: vat-remaining 0.561 with Renv+Remp vs 0.519 under full Rtotal). No load-bearing self-citation exists: the central machinery cites external works [17]-[21]; the authors' own citations ([26]-[29], [38], [39]) support only SNN compatibility and related work, and no uniqueness theorem is invoked. The Section 3 disclaimer ('the proposed intrinsic incentive mechanism is not predicated on enabling agents to comprehend human morality') is weighed in the authors' favor: the paper ultimately couches the mechanism as an incentive scheme, not as genuine moral understanding. The remaining objections — that Eq. 3 evaluates the agent's own action in the other's state and never uses s_others^{t+1}, so the proxy may not measure the other's welfare — are validity and generality risks, not circularity, and are not double-counted in this score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The framework rests on the AUP-style assumption that random reward functions probe an environment's attainable utility; on the domain assumption that imagination shares the real environment's transition dynamics; and on the paper-specific assumption that one agent's random-reward Q-functions estimate another agent's welfare. The hyperparameters α, β, and the number of imaginary spaces N are hand-chosen. No new entities are postulated.

free parameters (3)
  • alpha (weight on Rnse) = tested 1 to 20
    Controls the strength of the negative side-effect penalty. Chosen by hand, not fitted to data.
  • beta (weight on Remp) = tested 1 to 20
    Controls the strength of the empathy incentive. Chosen by hand, not fitted to data.
  • N (number of imaginary environments) = 30
    Number of random reward functions used to estimate Qi. Chosen by hand in Table A2.
assumptions (5)
  • domain assumption Transition functions in imaginary spaces are identical to the real environment.
    Section 4.1: 'the transition functions T in these imaginary spaces are identical to those in the real environment.' This is required for Qi to transfer to real decisions.
  • domain assumption Stepwise inaction state is an appropriate baseline for side-effect measurement.
    Section 4.2, adopted from AUP literature; chosen to avoid repeated penalties and to ensure inaction yields zero penalty.
  • ad hoc to paper One agent's random-reward Q-functions estimate another agent's well-being.
    Section 4.3: 'it is reasonable to directly use the same Qi to estimate the value of others' state... which reflects the essence of empathy.' This is the core ToM assumption, not proven.
  • domain assumption The non-standard max update in Eq. 1 converges to a useful value function.
    Eq. 1 uses max of old estimate and new target, a monotone operator. Convergence is not proven and is not a standard Q-learning update.
  • standard math The environment is a Markov Decision Process with global observation.
    MDP formulation in Section 4.1 and global observation in Appendix A.1 are standard and stated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Autonomous Alignment with Human Value on Altruism through Considerate Self-imagination and Theory of Mind." pith.science (2026). https://pith.science/paper/VHP62NS2

@misc{pith2026250100320,
  author       = {Pith},
  title        = {Pith review of: Autonomous Alignment with Human Value on Altruism through Considerate Self-imagination and Theory of Mind},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VHP62NS2}},
  note         = {Machine review of arXiv:2501.00320}
}
read the original abstract

With the widespread application of Artificial Intelligence (AI) in human society, enabling AI to autonomously align with human values has become a pressing issue to ensure its sustainable development and benefit to humanity. One of the most important aspects of aligning with human values is the necessity for agents to autonomously make altruistic, safe, and ethical decisions, considering and caring for human well-being. Current AI extremely pursues absolute superiority in certain tasks, remaining indifferent to the surrounding environment and other agents, which has led to numerous safety risks. Altruistic behavior in human society originates from humans' capacity for empathizing others, known as Theory of Mind (ToM), combined with predictive imaginative interactions before taking action to produce thoughtful and altruistic behaviors. Inspired by this, we are committed to endow agents with considerate self-imagination and ToM capabilities, driving them through implicit intrinsic motivations to autonomously align with human altruistic values. By integrating ToM within the imaginative space, agents keep an eye on the well-being of other agents in real time, proactively anticipate potential risks to themselves and others, and make thoughtful altruistic decisions that balance negative effects on the environment. The ancient Chinese story of Sima Guang Smashes the Vat illustrates the moral behavior of the young Sima Guang smashed a vat to save a child who had accidentally fallen into it, which is an excellent reference scenario for this paper. We design an experimental scenario similar to Sima Guang Smashes the Vat and its variants with different complexities, which reflects the trade-offs and comprehensive considerations between self-goals, altruistic rescue, and avoiding negative side effects.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Super Co-alignment of Human and AI for Sustainable Symbiotic Society

    cs.AI 2025-04 unverdicted novelty 4.0 of 10

    The authors propose 'Super Co-alignment', in which humans and superintelligent AI iteratively co-evolve shared values through external oversight and intrinsic empathy-based alignment.

Reference graph

Works this paper leans on

42 extracted references · 33 canonical work pages · cited by 1 Pith paper

  1. [1]

    arXiv preprint arXiv:1606.06565 (2016)

    Amodei, D., Olah, C., Steinhardt, J., Christiano, P., Schulman, J., Man´ e, D.: Concrete problems in ai safety. arXiv preprint arXiv:1606.06565 (2016)

    arXiv 2016

  2. [2]

    arXiv preprint arXiv:1711.09883 (2017)

    Leike, J., Martic, M., Krakovna, V., Ortega, P.A., Everitt, T., Lefrancq, A., Orseau, L., Legg, S.: Ai safety gridworlds. arXiv preprint arXiv:1711.09883 (2017)

    arXiv 2017

  3. [3]

    Patterns 5(5) (2024)

    Park, P.S., Goldstein, S., O’Gara, A., Chen, M., Hendrycks, D.: Ai deception: A survey of examples, risks, and potential solutions. Patterns 5(5) (2024)

    work page 2024

  4. [4]

    nature 575(7782), 350–354 (2019)

    Vinyals, O., Babuschkin, I., Czarnecki, W.M., Mathieu, M., Dudzik, A., Chung, J., Choi, D.H., Powell, R., Ewalds, T., Georgiev, P., et al.: Grandmaster level in starcraft ii using multi-agent reinforcement learning. nature 575(7782), 350–354 (2019)

    work page 2019

  5. [5]

    Science 365(6456), 885–890 (2019)

    Brown, N., Sandholm, T.: Superhuman ai for multiplayer poker. Science 365(6456), 885–890 (2019)

    work page 2019

  6. [6]

    Advances in neural information processing systems 30 (2017)

    Christiano, P.F., Leike, J., Brown, T., Martic, M., Legg, S., Amodei, D.: Deep reinforcement learning from human preferences. Advances in neural information processing systems 30 (2017)

    2017

  7. [7]

    Asimov, I.: I, Robot vol. 1. Spectra, New York (2004)

    work page 2004

  8. [8]

    Neuron 76(4), 677–694 (2012)

    Schacter, D.L., Addis, D.R., Hassabis, D., Martin, V.C., Spreng, R.N., Szpunar, K.K.: The future of memory: remembering, imagining, and the brain. Neuron 76(4), 677–694 (2012)

    work page 2012

Show all 42 references
  1. [9]

    Neuroimage 40(1), 398–407 (2008)

    D’Argembeau, A., Xue, G., Lu, Z.-L., Linden, M., Bechara, A.: Neural correlates of envisioning emotional events in the near and far future. Neuroimage 40(1), 398–407 (2008)

    2008

  2. [10]

    Philosophical Transactions of the Royal Society B: Biological Sciences 364(1521), 1263–1271 (2009)

    Hassabis, D., Maguire, E.A.: The construction system of the brain. Philosophical Transactions of the Royal Society B: Biological Sciences 364(1521), 1263–1271 (2009)

    2009

  3. [11]

    Advances in Psychological Science 23(3), 394 (2015)

    XU, X., YU, J., LEI, X.: Imagining the future: Cognitive processes and brain networks. Advances in Psychological Science 23(3), 394 (2015)

    2015

  4. [12]

    Social cognitive and affective neuroscience 7(1), 53–63 (2012)

    Sebastian, C.L., Fontaine, N.M., Bird, G., Blakemore, S.-J., De Brito, S.A., 20 McCrory, E.J., Viding, E.: Neural processing associated with cognitive and affec- tive theory of mind in adolescents and adults. Social cognitive and affective neuroscience 7(1), 53–63 (2012)

    2012

  5. [13]

    Developmental cognitive neuroscience 5, 25–39 (2013)

    Dennis, M., Simic, N., Bigler, E.D., Abildskov, T., Agostino, A., Taylor, H.G., Rubin, K., Vannatta, K., Gerhardt, C.A., Stancin, T., et al.: Cognitive, affec- tive, and conative theory of mind (tom) in children with traumatic brain injury. Developmental cognitive neuroscience...

    2013

  6. [14]

    In: IJCAI, pp

    Zhang, S., Durfee, E.H., Singh, S.: Minimax-regret querying on side effects for safe optimality in factored markov decision processes. In: IJCAI, pp. 4867–4873 (2018)

    2018

  7. [15]

    arXiv preprint arXiv:1805.00899 (2018)

    Irving, G., Christiano, P., Amodei, D.: Ai safety via debate. arXiv preprint arXiv:1805.00899 (2018)

    2018 arXiv

  8. [16]

    arXiv preprint arXiv:1705.10720 (2017)

    Armstrong, S., Levinstein, B.: Low impact artificial intelligences. arXiv preprint arXiv:1705.10720 (2017)

    2017 arXiv

  9. [17]

    arXiv preprint arXiv:1806.01186 (2018)

    Krakovna, V., Orseau, L., Kumar, R., Martic, M., Legg, S.: Penalizing side effects using stepwise relative reachability. arXiv preprint arXiv:1806.01186 (2018)

    2018 arXiv

  10. [18]

    Advances in Neural Information Processing Systems 33, 21406–21415 (2020)

    Turner, A., Ratzlaff, N., Tadepalli, P.: Avoiding side effects in complex envi- ronments. Advances in Neural Information Processing Systems 33, 21406–21415 (2020)

    2020

  11. [19]

    In: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, pp

    Turner, A.M., Hadfield-Menell, D., Tadepalli, P.: Conservative agency via attain- able utility preservation. In: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, pp. 385–391 (2020)

    2020

  12. [20]

    Advances in Neural Information Processing Systems 33, 19064–19074 (2020)

    Krakovna, V., Orseau, L., Ngo, R., Martic, M., Legg, S.: Avoiding side effects by considering future tasks. Advances in Neural Information Processing Systems 33, 19064–19074 (2020)

    2020

  13. [21]

    In: 2019 Workshop on Artificial Intelligence Safety, AISafety 2019, pp

    Bussmann, B., Heinerman, J., Lehman, J.: Towards empathic deep q-learning. In: 2019 Workshop on Artificial Intelligence Safety, AISafety 2019, pp. 1–7 (2019). CEUR-WS. org

    2019

  14. [22]

    In: Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems, pp

    Senadeera, M., Karimpanal, T.G., Gupta, S., Rana, S.: Sympathy-based rein- forcement learning agents. In: Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems, pp. 1164–1172 (2022)

    2022

  15. [23]

    In: Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems, pp

    Alizadeh Alamdari, P., Klassen, T.Q., Toro Icarte, R., McIlraith, S.A.: Be con- siderate: Avoiding negative side effects in reinforcement learning. In: Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems, pp. 18–26 (2022) 21

    2022

  16. [24]

    In: NeurIPS ML Safety Workshop (2022)

    Klassen, T.Q., Alamdari, P.A., McIlraith, S.A.: Epistemic side effects & avoiding them (sometimes). In: NeurIPS ML Safety Workshop (2022)

    2022

  17. [25]

    In: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pp

    Klassen, T.Q., Alamdari, P.A., McIlraith, S.A.: Epistemic side effects: An ai safety problem. In: Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, pp. 1797–1801 (2023)

    2023

  18. [26]

    Frontiers in Computational Neuroscience 16, 784967 (2022)

    Feng, H., Zeng, Y., Lu, E.: Brain-inspired affective empathy computational model and its application on altruistic rescue task. Frontiers in Computational Neuroscience 16, 784967 (2022)

    2022

  19. [27]

    arXiv preprint arXiv:2410.21882 (2024)

    Zhao, F., Feng, H., Tong, H., Han, Z., Lu, E., Sun, Y., Zeng, Y.: Building altruis- tic and moral ai agent with brain-inspired affective empathy mechanisms. arXiv preprint arXiv:2410.21882 (2024)

    2024

  20. [28]

    Frontiers in neuroscience 16, 753900 (2022)

    Zhao, Z., Lu, E., Zhao, F., Zeng, Y., Zhao, Y.: A brain-inspired theory of mind spiking neural network for reducing safety risks of other agents. Frontiers in neuroscience 16, 753900 (2022)

    2022

  21. [29]

    Patterns 4(8) (2023)

    Zhao, Z., Zhao, F., Zhao, Y., Zeng, Y., Sun, Y.: A brain-inspired theory of mind spiking neural network improves multi-agent cooperation and competition. Patterns 4(8) (2023)

    2023

  22. [30]

    arXiv preprint arXiv:1910.01708 7(1), 2 (2019)

    Ray, A., Achiam, J., Amodei, D.: Benchmarking safe exploration in deep reinforcement learning. arXiv preprint arXiv:1910.01708 7(1), 2 (2019)

    2019 arXiv

  23. [31]

    Advances in Neural Information Processing Systems 36 (2023)

    Ji, J., Zhang, B., Zhou, J., Pan, X., Huang, W., Sun, R., Geng, Y., Zhong, Y., Dai, J., Yang, Y.: Safety gymnasium: A unified safe reinforcement learning benchmark. Advances in Neural Information Processing Systems 36 (2023)

    2023

  24. [32]

    arXiv preprint arXiv:1912.01217 (2019)

    Wainwright, C.L., Eckersley, P.: Safelife 1.0: Exploring side effects in complex environments. arXiv preprint arXiv:1912.01217 (2019)

    2019 arXiv

  25. [33]

    arXiv preprint arXiv:2003.11881 (2020)

    Dulac-Arnold, G., Levine, N., Mankowitz, D.J., Li, J., Paduraru, C., Gowal, S., Hester, T.: An empirical investigation of the challenges of real-world reinforcement learning. arXiv preprint arXiv:2003.11881 (2020)

    2020 arXiv

  26. [34]

    nature 518(7540), 529–533 (2015)

    Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A.A., Veness, J., Bellemare, M.G., Graves, A., Riedmiller, M., Fidjeland, A.K., Ostrovski, G., et al.: Human-level control through deep reinforcement learning. nature 518(7540), 529–533 (2015)

    2015

  27. [35]

    Neural networks 10(9), 1659–1671 (1997)

    Maass, W.: Networks of spiking neurons: the third generation of neural network models. Neural networks 10(9), 1659–1671 (1997)

    1997

  28. [36]

    In: Proceedings of the 27th International Conference on Machine Learning (ICML-10), pp

    Nair, V., Hinton, G.E.: Rectified linear units improve restricted boltzmann machines. In: Proceedings of the 27th International Conference on Machine Learning (ICML-10), pp. 807–814 (2010) 22

    2010

  29. [37]

    MIT press, ??? (2005)

    Dayan, P., Abbott, L.F.: Theoretical Neuroscience: Computational and Mathe- matical Modeling of Neural Systems. MIT press, ??? (2005)

    2005

  30. [38]

    Patterns 4(8) (2023)

    Zeng, Y., Zhao, D., Zhao, F., Shen, G., Dong, Y., Lu, E., Zhang, Q., Sun, Y., Liang, Q., Zhao, Y., et al.: Braincog: A spiking neural network based, brain- inspired cognitive intelligence engine for brain-inspired ai and brain simulation. Patterns 4(8) (2023)

    2023

  31. [39]

    Patterns 3(6) (2022)

    Shen, G., Zhao, D., Zeng, Y.: Backpropagation with biologically plausible spa- tiotemporal adjustment for training deep spiking neural networks. Patterns 3(6) (2022)

    2022

  32. [40]

    Nature Computational Science, 1–18 (2024)

    Wu, X., Ren, X., Liu, C., Zhang, H.: The motive cocktail in altruistic behaviors. Nature Computational Science, 1–18 (2024)

    2024

  33. [41]

    Proceedings of the National Academy of Sciences 121(17), 2314590121 (2024)

    Jin, K., Wu, J., Zhang, R., Zhang, S., Wu, X., Wu, T., Gu, R., Liu, C.: Observing heroic behavior and its influencing factors in immersive virtual environments. Proceedings of the National Academy of Sciences 121(17), 2314590121 (2024)

    2024

  34. [42]

    Frontiers in neuroscience 11, 682 (2017) 23

    Rueckauer, B., Lungu, I.-A., Hu, Y., Pfeiffer, M., Liu, S.-C.: Conversion of continuous-valued deep networks to efficient event-driven networks for image classification. Frontiers in neuroscience 11, 682 (2017) 23

    2017

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.