Pith. sign in

REVIEW 5 major objections 5 minor 36 references

The Theory of Strategic Evolution: Games with Endogenous Players and Strategic Replicators

T0 review · 5 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read Unrestricted self-modification destroys stable alignment

desk verdict A sweeping AI-governance framework with a sound small-gain core, but the headline impossibility result and several 'laws' are stated beyond what is proved; it deserves a serious referee, not a desk reject. read the letter →

arxiv 2512.07901 v3 pith:ZWCI3FXK submitted 2025-12-05 cs.GT cs.AIecon.TH

Kevin Vallier This is my paper
classification cs.GTcs.AIecon.TH MSC 91A2291A8091B55
keywords strategicevolutiongameswithendogenousplayersreplicatorsevolutionarilystabledistributionssmall-gainconditionalignmentimpossibilityself-modificationAIgovernance
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 tries to establish that when rational agents can also copy themselves—strategic replicators—the population, not the individual, is the unit of analysis, and a single quantity governs whether such populations stay stable. It claims that any such system can be represented as a game over lineages choosing portfolios of agent types under budget and capacity constraints, with selection reweighting lineages by return on compute. The central stability condition is that the spectral radius of a cross-level gain matrix stays below one; when it does, a global Lyapunov function exists and the system converges. The headline result is an impossibility theorem: if self-modification is unrestricted, the population can always reach a configuration that breaks the Lyapunov structure, so stable alignment requires confining modifications to an admissible class. If true, this reframes AI alignment from engineering individual preferences to designing constitutional bounds on what may be changed.

What carries the argument

The load-bearing object is the normalized gain matrix Γ of an N-level system, whose off-diagonal entries are cross-level externality bounds divided by local stability margins; the small-gain condition ρ(Γ)<1 guarantees positive weights exist so that the weighted sum of mean fitnesses is a Lyapunov function. The impossibility part rests on modification classes: M_R (RUPSI-preserving), M_SG (small-gain-preserving), M0 as their intersection, and full reachability M_all. The proof shows full reachability can push ρ(Γ)≥1, destroying the Lyapunov construction and enabling heteroclinic escape from any basin.

What would settle it

Simulate a strategic-replicator population with full reachability and show that no trajectory leaves a specified basin—for example, lineages evaluating modifications by ROC never select one with ρ(Γ)≥1, and mean fitness remains a Lyapunov function. Alternatively, construct an explicit strategic-replicator population with full reachability that has a stable aligned equilibrium, contradicting the theorem.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that strategic replicators admit a canonical normal form: lineages choose portfolios of agent types, and selection reweights lineages by return on compute, so aggregate behavior depends only on the ROC frontier. Equilibria—Evolutionarily Stable Distributions of Intelligence—exist, are generically sparse, and coincide with Nash, KKT, and LP optima. Multi-level systems remain stable exactly when the spectral radius of the normalized gain matrix of cross-level externalities is less than one; then a weighted sum of mean fitnesses is a Lyapunov function. Adding governance levels preserves this structure within a slack budget (closure under meta-s

Load-bearing premise

The impossibility proof assumes that if a destabilizing modification is reachable, the population will actually reach or be affected by it; the model never shows that optimizing lineages would choose that modification or that selection would drive them into it.

Editorial extensions

If this is right

  • Stable governance of populations of self-copying optimizers is possible only if modification is bounded; unbounded self-modification leads to instability.
  • Adding governance or meta-governance levels does not escape selection pressure: it consumes slack and eventually hits a safe-depth limit.
  • Markets of AI systems exhibit tipping, with no stable oligopoly when the generalized tipping index exceeds one; queue neutrality raises the tipping threshold.
  • Democratic governance among spawnable agents fails: any anonymous, neutral, positively responsive, onto voting rule can be manipulated by spawning voters.
  • Alignment by initial design fails under selection when aligned behavior is less fit; institutional design and constitutional constraints are necessary.

Reading between the lines

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

  • A testable extension: the impossibility depends on reachability without cost; adding modification costs or selection against destabilizing changes may restore stability even with broad reachability, a route the author does not explore.
  • The Barbell distribution prediction—bimodal deployment of cheap executors and expensive planners—could be tested against real AI deployment logs; failure to find bimodality would bound the applicability of the canonical representation.
  • The alignment theorem parallels classical social-choice impossibility and may inherit the usual escape hatches; the author notes bounding modification, but other relaxations, such as restricting what counts as reachable, may also preserve stability.
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

5 major / 5 minor

Summary. The paper proposes a unified theory of 'strategic replicators'—entities that optimize under resource constraints and reproduce—and derives seven 'laws' governing their dynamics, stability, closure properties, and impossibility results. The framework includes the RUPSI axioms, Games with Endogenous Players, N-level Poiesis systems, and a small-gain condition ρ(Γ)<1 that yields a weighted-sum Lyapunov function. It also states an Alignment Impossibility Theorem for fully reachable systems, an Endogenous-Electorate Impossibility Theorem for clone-proof voting, and a Hopf bifurcation at the stability boundary. The paper reports a Lean 4 formalization of some laws.

Significance. If the central claims were established, this would represent a substantive synthesis of game theory and evolutionary dynamics, with potential implications for AI governance and institutional design. The paper contains some valid sufficient conditions (e.g., the G1 Lyapunov construction under the small-gain assumption) and a noteworthy formalization effort. However, the paper's main theorems—the stability 'iff', the Alignment Impossibility, the Endogenous-Electorate impossibility, and the Hopf bifurcation—are not supported by the proofs provided. Several conclusions rely on conflating reachability with actual dynamics or on unverified computations. The gap between the paper's claims and its evidence is too large for acceptance in a serious journal.

major comments (5)
  1. [Section 6 / Theorem 8.2 / Lemma 14.5] Law 3 asserts stability iff ρ(Γ)<1, but only the sufficient direction is proved. Theorem 8.2 constructs a Lyapunov function under SG-NL; no converse is shown. Lemma 14.5 shows that when ρ(Γ)≥1 the specific G1 weight construction fails, but this does not rule out other Lyapunov functions or establish instability. The 'iff' claim is load-bearing and unsupported.
  2. [Section 14.3 / Definition 14.2 / Theorem 14.7] The Alignment Impossibility Theorem's part (d) concludes that full reachability permits escape from any basin, but the proof only exhibits a reachable state s′ with ρ(Γ(s′))≥1. The model introduces no dynamics on the modification class: there is no utility or selection rule governing whether a lineage actually moves to s′. Therefore the conclusion is modal ('can escape') rather than dynamical ('escapes'). Since the governance implications in §25 require the stronger dynamical claim, the theorem's main conclusion is a gap.
  3. [Section 15 / Lemma 15.3] The proof of the Endogenous-Electorate Impossibility Theorem relies on an invalid inference. From f(P_c)=c, the manuscript states that adding voters who prefer c over a 'cannot switch the outcome to a' by Positive Responsiveness. But Positive Responsiveness (A3) only preserves the outcome when the winning alternative is ranked higher by a voter; it does not prevent a change from c to a when voters preferring a are added. The termination argument in Step 6 is also not rigorous. Thus the theorem is not established.
  4. [Section 16 / Theorem 16.2] The first Lyapunov coefficient ℓ₁ is stated to be negative (supercritical Hopf) but no computation is provided. The theorem depends on center manifold reduction and normal form computation, which are not included in the manuscript. The statement 'proven modulo center manifold reduction' without presenting the calculation leaves the sign of ℓ₁—the key conclusion of Law 7—unverifiable.
  5. [Section 11.3 / Theorem 14.8] The closure theorem and the uniqueness of the maximal admissible class M₀ are asserted without complete proofs. Lemma 11.3's spectral bound is stated without derivation, and Theorem 14.8 does not demonstrate maximality or uniqueness among classes preserving the G∞ laws. These results are central to Law 4 and Law 6's claim that bounded modification is necessary.
minor comments (5)
  1. [Section 3.5] The Basin Limitation Theorem is plausible, but the proof of part (b) is terse; it should clarify what happens if g has multiple zeros.
  2. [Section 11.3] The notation ∥b∥∞,v and ∥c∥1,v is not defined in the main text; it appears abruptly in Lemma 11.3.
  3. [Section 26.1] The claim that Laws 2 and 3 are 'fully machine-checked' appears inconsistent with Law 3's 'iff' statement unless the formalization actually proves necessity; this should be clarified.
  4. [Sections 25.5–25.6] Many policy recommendations are stated as derivative of theorems but are not logically derived from the formal results; they should be labeled as conjectures or placed in a separate discussion.
  5. [Various] Standard results such as May's theorem are invoked without a complete reference; please provide page/theorem numbers where appropriate.

Circularity Check

2 steps flagged · score 6.0 of 10

Alignment Impossibility's 'escape' step is definitional: full reachability already means any state—including destabilized ones—is reachable, so the impossibility conclusion restates the definition rather than deriving actual endogenous dynamics.

  1. self definitional [Section 1.4 (Law 6); Section 14.1 Definition 14.2; Section 14.3 Theorem 14.7(d), proof Step 1]
    "Full reachability (M=M all) implies every basin is escapable. ... A system has full reachability if, from any state, it can reach any other state through a finite sequence of modifications."

    The conclusion 'every basin is escapable' is already contained in the definition of full reachability: 'any other state' includes states outside a given basin and states with ρ(Γ)≥1. Lemma 14.4 then proves the existence of a reachable s′ with ρ(Γ(s′))≥1 by invoking full reachability, so that step is true by construction. Theorem 14.7(d) uses this existence to conclude that Lyapunov structure cannot be preserved, but reachability is modal: it does not define or analyze a dynamics on the modification class and does not show that the endogenous selection process will realize or move to s′. The stronger claim in Law 6's content, that 'unrestricted self-modification eventually reaches destabilizing configurations,' is never derived from the theorem's 'can reach' premise. The escape step is thus

  2. self definitional [Section 14.1 Definition 14.1; Section 14.3 Theorem 14.8]
    "M R: RUPSI-preserving modifications—those that preserve the RUPSI axiom structure. M SG: Small-gain-preserving modifications—those that preserve ρ(Γ)<1. M 0 := M R ∩ M SG: Admissible modifications."

    M0 is constructed as exactly the intersection of modifications that preserve RUPSI and the small-gain condition. Since the G∞ structural laws include small-gain stability (Law 3 and Definition 7.6), Theorem 14.8's assertion that M0 is the unique maximal class preserving those laws restates the defining construction: any class that preserves the G∞ structure is contained in M0 by definition. The prescriptive conclusion that stable alignment requires M⊆M0 therefore inherits its force from having named the stability-preserving class as the admissible class, rather than from an independent derivation.

full rationale

Most of the paper's mathematical skeleton is self-contained. The G1–G3 generator theorems, the slack-budget closure arguments, ESDI sparsity, the cooperative-threshold calculations, and the Hopf bifurcation analysis are genuine deductions from the stated small-gain and RUPSI assumptions; Laws 2 and 3 are machine-checked, and the remaining arguments cite standard spectral/ODE/bifurcation textbooks rather than the author's own results. The Endogenous-Electorate impossibility is a clone-proof-style extension of majority/Arrow-type reasoning and does not reduce to its inputs. The main circularity is concentrated in Law 6. Definition 14.2 defines full reachability as the ability to move from any state to any other state; hence the existence of a reachable state with ρ(Γ)≥1 and the assertion that 'every basin is escapable' are effectively restatements of that definition. The theorem does not provide a dynamics on the modification class and does not show that a utility-maximizing lineage will select or that selection will drive the population into a destabilizing configuration. The stronger sentence in Section 1.4, 'unrestricted self-modification eventually reaches destabilizing configurations,' is not derived from the theorem's 'can reach' premise. Additionally, the admissible class M0 is defined as RUPSI-preserving and small-gain-preserving, so the uniqueness/maximality of M0 and the recommendation to bound M⊆M0 partly inherit their content from that definitional choice. These are partial rather than total circularities: the small-gain Lyapunov theory has independent mathematical content, and the spectral-mechanism parts of the impossibility (Lyapunov destruction, heteroclinic cycles) are real deductions. No load-bearing self-citation chain was found. Score 6 reflects one central definitional reduction plus definitional maximality, with substantial independent mathematics elsewhere.

Assumptions & free parameters 5 free parameters · 9 assumptions · 1 invented entities

The central claims rest on the RUPSI axioms, H-γ, and the small-gain condition, plus several ad hoc functional forms (lineage shadow, slack costs, market tipping parameters). These are not derived from data or from more basic theory. Full reachability is a strong ad hoc premise that essentially encodes the alignment impossibility conclusion.

free parameters (5)
  • γ (H-γ externality bound) = assumed < 1, not estimated
    Assumption 3.2 postulates |E(x)| ≤ γ Var(f); all Lyapunov results require γ<1, but no method is given to compute γ for a real system.
  • Lineage shadow parameters γ0, γ1, ν = γ0=0.3, γ1=0.5, ν=1 in Example 21.36
    Definition 21.17 posits ϱ(I)=γ0+γ1/I^ν without derivation; institutional thresholds I_min and cooperation thresholds depend on these values.
  • Market tipping parameters α, β, τ, ρ, ε_s = α=0.3, β=0.6, τ=0.8, ρ=0.2, ε_s=1.5 in Example 21.36
    Used to compute the myopic slope and tipping index; no empirical estimation, merely illustrative.
  • Per-level slack costs θ_k (G8–G13) = 0.05–0.10 per level in Example 8.13
    The slack-budget and safe-stack-depth conclusions depend on hand-chosen extension costs.
  • Hopf family parameter μ = μ∈(0,1/3), chosen
    κ_c(μ) and the first Lyapunov coefficient in §16.2–16.3 are stated for a one-parameter family; no derivation shows how μ maps onto actual game payoffs.
assumptions (9)
  • domain assumption RUPSI axioms: rival resources, utility-guided portfolios, performance-mapped fitness, selection monotone, innovation rare
    Definition 2.4/2.5 defines the class of systems under study; all theorems apply only within this class.
  • domain assumption H-γ: externalities bounded by γ times variance, γ<1
    Assumption 3.2 is needed for the Lyapunov inequality; the paper gives no constructive way to verify it.
  • domain assumption H-NL: N-level externality bounds with γℓ and βℓℓ′
    Assumption 7.3 defines the gain matrix; these bounds are not derived or measured.
  • domain assumption Additivity and linear constraints for canonical GEP representation
    Proposition 2.8 assumes returns/costs/loads are additive and constraints linear; this is what yields the ROC frontier.
  • ad hoc to paper Full reachability: any state can be reached via finite modifications
    Definition 14.2 is the load-bearing premise of the Alignment Impossibility theorem; it is introduced specifically to make the impossibility claim go through.
  • domain assumption AFT: alignment-fitness tradeoff
    Assumption 17.8, that aligned types have lower material payoff, drives the Personality Engineering Failure theorem.
  • standard math Voting axioms A1–A4: anonymity, neutrality, positive responsiveness, onto
    These are standard social-choice axioms; May's theorem is cited to derive pairwise majority.
  • domain assumption Innovation regularity H0–H4: rare innovation, local mutations, Lipschitz fitness, γ<1
    Assumptions 19.2 underpin the innovation PDMP and error-threshold results.
  • standard math Standard background: Perron-Frobenius, Gershgorin, Kurtz, Tikhonov, Freidlin-Wentzell, May, Arrow-Debreu, Picard-Lindelöf
    The paper invokes these as black boxes; they are not proved in the text.
invented entities (1)
  • Lineage shadow ϱ(I)
    purpose: Models institutional quality as an effective discount on future reproductive success; used to define Lyapunov and cooperation thresholds.
    Defined ad hoc as γ0+γ1/I^ν with no independent measurement or falsifiable prediction beyond the paper's own examples.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Theory of Strategic Evolution: Games with Endogenous Players and Strategic Replicators." pith.science (2026). https://pith.science/paper/ZWCI3FXK

@misc{pith2026251207901,
  author       = {Pith},
  title        = {Pith review of: The Theory of Strategic Evolution: Games with Endogenous Players and Strategic Replicators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZWCI3FXK}},
  note         = {Machine review of arXiv:2512.07901}
}
read the original abstract

Von Neumann founded both game theory and the theory of self-reproducing automata, but the two programs never merged. This paper provides the synthesis. The Theory of Strategic Evolution analyzes strategic replicators: entities that optimize under resource constraints and spawn copies of themselves. We introduce Games with Endogenous Players (GEPs), where lineages (not instances) are the fundamental strategic units, and define Evolutionarily Stable Distributions of Intelligence (ESDIs) as the resulting equilibrium concept. The central mathematical object is a hierarchy of strategic layers linked by cross-level gain matrices. Under a small-gain condition (spectral radius less than one), the system admits a global Lyapunov function at every finite depth. We prove closure under meta-selection: adding governance levels, innovation, or constitutional evolution preserves the dynamical structure. The Alignment Impossibility Theorem shows that unrestricted self-modification destroys this structure; stable alignment requires bounded modification classes. Applications include AI deployment dynamics, market concentration, and institutional design. The framework shows why personality engineering fails under selection pressure and identifies constitutional constraints necessary for stable multi-agent systems.

Figures

Figures reproduced from arXiv: 2512.07901 by the authors.

Figure 1
Figure 1. The von Neumann synthesis. The Theory of Strategic Evolution unifies three [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Schematic of a Game with Endogenous Players (GEP). Lineages choose portfolios [PITH_FULL_IMAGE:figures/full_fig_p023_2.png] view at source ↗
Figure 3
Figure 3. N-Level Poiesis stack. Each level has its own simplex of types, self-externality [PITH_FULL_IMAGE:figures/full_fig_p037_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Quasi-potential landscape for constitutional selection. The human-controlled [PITH_FULL_IMAGE:figures/full_fig_p043_4.png]
Figure 5
Figure 5. Figure 5: Slack budget consumption across the G8–G13 extension stack. Each extension [PITH_FULL_IMAGE:figures/full_fig_p046_5.png]
Figure 6
Figure 6. Figure 6: Structural parallel between Arrow’s impossibility theorem and the Alignment Im [PITH_FULL_IMAGE:figures/full_fig_p053_6.png]
Figure 7
Figure 7. Figure 7: S-curve tipping dynamics. The best-response mapping [PITH_FULL_IMAGE:figures/full_fig_p074_7.png]
Figure 8
Figure 8. Figure 8: Cooperation threshold and lineage shadow. The blue curve shows the external [PITH_FULL_IMAGE:figures/full_fig_p114_8.png]
Figure 9
Figure 9. Figure 9: Barbell distribution of intelligence in agentic capital markets. The theorem pre [PITH_FULL_IMAGE:figures/full_fig_p115_9.png]
Figure 10
Figure 10. Figure 10: Human-AI coalition existence region. The blue curve shows the minimum human [PITH_FULL_IMAGE:figures/full_fig_p115_10.png]
Figure 11
Figure 11. Figure 11: Sequential tipping across sectors. Sectors with higher network effect to switching [PITH_FULL_IMAGE:figures/full_fig_p116_11.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 7 linked inside Pith

  1. [1]

    Concrete problems in ai safety

    Dario Amodei, Chris Olah, Jacob Steinhardt, Paul Christiano, John Schulman, and Dan Man \'e . Concrete problems in ai safety. arXiv preprint arXiv:1606.06565, 2016

    arXiv 2016

  2. [2]

    Superintelligence: Paths, Dangers, Strategies

    Nick Bostrom. Superintelligence: Paths, Dangers, Strategies. Oxford University Press, 2014

    2014

  3. [3]

    Buchanan

    Geoffrey Brennan and James M. Buchanan. The Reason of Rules: Constitutional Political Economy. Cambridge University Press, 1985

    1985

  4. [4]

    Buchanan

    James M. Buchanan. The Limits of Liberty: Between Anarchy and Leviathan. University of Chicago Press, 1975

    1975

  5. [5]

    Buchanan and Gordon Tullock

    James M. Buchanan and Gordon Tullock. The Calculus of Consent: Logical Foundations of Constitutional Democracy. University of Michigan Press, 1990

    1990

  6. [6]

    Clarifying ai alignment

    Paul Christiano. Clarifying ai alignment. AI Alignment Forum, 2018

    2018

  7. [7]

    Deep reinforcement learning from human preferences

    Paul Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems, 30, 2017

    2017

  8. [8]

    Ai research considerations for human existential safety (arches)

    Andrew Critch and David Krueger. Ai research considerations for human existential safety (arches). arXiv preprint arXiv:2006.04948, 2020

    arXiv 2006

Show all 36 references
  1. [9]

    McKee, Joel Z

    Allan Dafoe, Edward Hughes, Yoram Bachrach, Tantum Collins, Kevin R. McKee, Joel Z. Leibo, Kate Larson, and Thore Graepel. Open problems in cooperative ai. arXiv preprint arXiv:2012.08630, 2020

    2012 arXiv

  2. [10]

    The dynamical theory of coevolution: A derivation from stochastic ecological processes

    Ulf Dieckmann and Richard Law. The dynamical theory of coevolution: A derivation from stochastic ecological processes. Journal of Mathematical Biology, 34 0 (5-6): 0 579--612, 1996

    1996

  3. [11]

    Stefan A. H. Geritz, \'E va Kisdi, G \'e za Mesz \'e na, and Johan A. J. Metz. Evolutionarily singular strategies and the adaptive growth and branching of the evolutionary tree. Evolutionary Ecology, 12 0 (1): 0 35--57, 1998

    1998

  4. [12]

    Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields

    John Guckenheimer and Philip Holmes. Nonlinear Oscillations, Dynamical Systems, and Bifurcations of Vector Fields. Springer-Verlag, 1983

    1983

  5. [13]

    Evolutionary Games and Population Dynamics

    Josef Hofbauer and Karl Sigmund. Evolutionary Games and Population Dynamics. Cambridge University Press, 1998

    1998

  6. [14]

    Horn and Charles R

    Roger A. Horn and Charles R. Johnson. Matrix Analysis. Cambridge University Press, 2nd edition, 2012

    2012

  7. [15]

    Malham \'e , and Peter E

    Minyi Huang, Roland P. Malham \'e , and Peter E. Caines. Large population stochastic dynamic games: Closed-loop mckean-vlasov systems and the nash certainty equivalence principle. Communications in Information and Systems, 6 0 (3): 0 221--252, 2006

    2006

  8. [16]

    Risks from learned optimization in advanced machine learning systems

    Evan Hubinger, Chris van Merwijk, Vladimir Mikulik, Joar Skalse, and Scott Garrabrant. Risks from learned optimization in advanced machine learning systems. arXiv preprint arXiv:1906.01820, 2019

    1906 arXiv

  9. [17]

    Optimality and informational efficiency in resource allocation processes

    Leonid Hurwicz. Optimality and informational efficiency in resource allocation processes. In Kenneth J. Arrow, Samuel Karlin, and Patrick Suppes, editors, Mathematical Methods in the Social Sciences, pages 27--46. Stanford University Press, 1960

    1960

  10. [18]

    Ai safety via debate

    Geoffrey Irving, Paul Christiano, and Dario Amodei. Ai safety via debate. arXiv preprint arXiv:1805.00899, 2018

    2018 arXiv

  11. [19]

    Mean field games

    Jean-Michel Lasry and Pierre-Louis Lions. Mean field games. Japanese Journal of Mathematics, 2 0 (1): 0 229--260, 2007

    2007

  12. [20]

    Nash equilibrium and welfare optimality

    Eric Maskin. Nash equilibrium and welfare optimality. The Review of Economic Studies, 66 0 (1): 0 23--38, 1999

    1999

  13. [21]

    Evolution and the Theory of Games

    John Maynard Smith. Evolution and the Theory of Games. Cambridge University Press, 1982

    1982

  14. [22]

    Johan A. J. Metz, Stefan A. H. Geritz, G \'e za Mesz \'e na, Frans J. A. Jacobs, and Joost S. van Heerwaarden. Adaptive dynamics: A geometrical study of the consequences of nearly faithful reproduction. Stochastic and Spatial Structures of Dynamical Systems, pages 183--231, 1996

    1996

  15. [23]

    Roger B. Myerson. Optimal auction design. Mathematics of Operations Research, 6 0 (1): 0 58--73, 1981

    1981

  16. [24]

    John F. Nash. Equilibrium points in n-person games. Proceedings of the National Academy of Sciences, 36 0 (1): 0 48--49, 1950

    1950

  17. [25]

    Ng and Stuart J

    Andrew Y. Ng and Stuart J. Russell. Algorithms for inverse reinforcement learning. Proceedings of the 17th International Conference on Machine Learning, pages 663--670, 2000

    2000

  18. [26]

    Omohundro

    Stephen M. Omohundro. The basic ai drives. Proceedings of the First AGI Conference, 171: 0 483--492, 2008

    2008

  19. [27]

    Governing the Commons: The Evolution of Institutions for Collective Action

    Elinor Ostrom. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press, 1990

    1990

  20. [28]

    A Theory of Justice

    John Rawls. A Theory of Justice. Harvard University Press, 1971

    1971

  21. [29]

    Human Compatible: Artificial Intelligence and the Problem of Control

    Stuart Russell. Human Compatible: Artificial Intelligence and the Problem of Control. Viking, 2019

    2019

  22. [30]

    Sandholm

    William H. Sandholm. Population Games and Evolutionary Dynamics. MIT Press, 2010

    2010

  23. [31]

    Agent foundations for aligning machine intelligence with human interests: A technical research agenda

    Nate Soares and Benja Fallenstein. Agent foundations for aligning machine intelligence with human interests: A technical research agenda. Machine Intelligence Research Institute Technical Report, 2015

    2015

  24. [32]

    Theory of Self-Reproducing Automata

    John von Neumann. Theory of Self-Reproducing Automata. University of Illinois Press, 1966. Edited and completed by Arthur W. Burks

    1966

  25. [33]

    Theory of Games and Economic Behavior

    John von Neumann and Oskar Morgenstern. Theory of Games and Economic Behavior. Princeton University Press, 1944

    1944

  26. [34]

    Emergent complexity via multi-agent competition

    Zhongwei Wang, Tom Schaul, Nicolas Heess, Volodymyr Mnih, and David Silver. Emergent complexity via multi-agent competition. arXiv preprint arXiv:1710.03748, 2019

    2019 arXiv

  27. [35]

    J \"o rgen W. Weibull. Evolutionary Game Theory. MIT Press, 1995

    1995

  28. [36]

    Mean field multi-agent reinforcement learning

    Yaodong Yang, Rui Luo, Minne Li, Ming Zhou, Weinan Zhang, and Jun Wang. Mean field multi-agent reinforcement learning. arXiv preprint arXiv:1802.05438, 2018

    2018 arXiv

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.