Pith. sign in

REVIEW 2 major objections 4 minor 30 references

The paper argues that the most serious AI risks are not failures of individual models but emergent properties of complex sociotechnical systems, and that they should be governed as collective action problems and externalities rather than as

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 15:45 UTC pith:4T4TRGUH

load-bearing objection A careful, well-structured conceptual synthesis of AI systemic risk as collective action plus externalities; the policy conclusion leans on an untested emergence premise, but the framework is still worth referee time. the 2 major comments →

arxiv 2607.18170 v1 pith:4T4TRGUH submitted 2026-07-20 cs.CY

An approach to systemic risks of AI through the lens of emergence, collective action problems, and externalities

Carsten Orwat , Lucas Staab , Alexandros Gazos This is my paper
classification cs.CY
keywords systemic riskartificial intelligenceemergencecollective action problemsexternalitiesfundamental rightsalgorithmic monoculturefeedback loops
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper is trying to establish that systemic risks from AI cannot be understood by examining individual AI systems one at a time. Its central claim is that societal goods like fundamental rights, democracy, and sustainability are emergent properties of complex systems, and AI creates systemic risk when micro-level interactions—between people, organisations, and AI systems—threaten those macro-level goods. The authors propose that these risks be analysed as collective action problems and complex externalities, amplified by feedback loops, market concentration, algorithmic monocultures, information asymmetries, and institutional weaknesses. If the framing is right, conventional risk assessment focused on model capabilities misses the main sources of harm, and governance must target system structure and incentives, not just frontier-model safety.

Core claim

In the paper's own terms, the core discovery is that systemic risks of AI are best conceptualised not as low-probability, high-impact tail risks from technical failure or malicious use, but as the interplay of risk phenomena—collective action problems and externalities in complex situations—that threaten or impede the realisation and protection of emergent properties at the societal or global level. Harms such as systemic discrimination, erosion of privacy, democratic degradation, and environmental damage emerge from the interaction of many actors and systems, so they cannot be reduced to the sum of individual model failures. The paper catalogues the phenomena that drive these dynamics—infor

What carries the argument

The central object is the concept of 'emergent properties': macro-level societal goods (fundamental rights, democracy, sustainability) that arise from the interactions of micro-level elements in a complex sociotechnical system and cannot be reduced to those elements' individual features. The mechanism carrying the argument is the pairing of 'collective action problems' (free-riding and coordination failures that undermine common goods) with 'complex externalities' (harms caused by many interacting actors whose individual contributions cannot be disentangled). These two concepts do the work of explaining how ordinary AI deployments, each apparently benign, can combine into systemic harms.

Load-bearing premise

The load-bearing premise is that AI-driven societal harms are genuinely emergent—that macro-level damage cannot be predicted from or reduced to individual model behaviours, and that non-linear amplification actually happens at societal scale; if harms are merely additive, the macro-level framing loses its distinctive force.

What would settle it

A concrete falsifier would be a large-scale empirical study showing that aggregate societal harms (e.g., discriminatory outcomes across a labour market) are well predicted by the simple sum of individual AI-system error rates, with no interaction, feedback, or cascade effects beyond what model-by-model assessment would anticipate. A specific version: if measured bias in real-world hiring after many firms adopt similar tools equals the average single-tool bias times the number of firms, with no amplification from training-data feedback or network effects, the emergence claim is weakened.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Risk assessment that evaluates individual AI models in isolation will systematically miss systemic harms; assessment must reconstruct micro-macro interaction pathways.
  • Responsibility cannot rest only on general-purpose AI providers; distributed actors and public institutions must share assessment and mitigation, with attention to structural dominance.
  • Market concentration and algorithmic monoculture turn private decisions into collective risks, so competition and ecosystem diversity become risk-governance issues.
  • Feedback loops (e.g., model collapse, fairness feedback loops, predictive policing) can push systems past thresholds, so monitoring should watch for non-linear amplification, not just accumulating incidents.
  • Governance gaps and perverse incentives in institutions can themselves cause systemic risks, so regulatory design is part of the risk system.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If emergence is real, measuring systemic risk requires new empirical methods—such as detecting early-warning signs of macro-state shifts (tipping points) rather than counting individual incidents; this is an untested extension.
  • The framework suggests a testable hypothesis: that societal-level indicators (e.g., aggregate discrimination or privacy erosion) will show non-linear jumps when AI adoption crosses certain thresholds, similar to phase transitions.
  • A corollary the authors leave implicit is that the same emergence logic applies to positive goals—well-designed polycentric governance could produce beneficial emergent properties, so interventions should target feedback architecture, not just prohibitions.
  • The paper's critique implies that any regulation keyed only to model capabilities will be under-inclusive; systemic risks from many small models interacting could be comparable to those from a few frontier models.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. This paper proposes a conceptual framework for systemic risks of AI based on complexity and emergence. It argues that existing definitions, especially the EU AI Act's focus on GPAI high-impact capabilities, neglect the systemic nature of harms that arise from interactions among many actors and technical components. The authors position societal goods—fundamental rights, democracy, sustainability—as emergent macro-level properties, and characterize systemic AI risks as the result of collective action problems and complex externalities. They catalog numerous contributing phenomena: informational emergence, proxy discrimination, integration and coordination failures in AI value chains, market concentration and network effects, algorithmic monoculture and outcome homogenisation, feedback loops, rebound effects, information asymmetries, normative ambiguities, and institutional deficits. Section 6 discusses how these phenomena interact and accumulate, with non-linearities and path dependencies; Section 7 lists open research questions; Section 8 concludes that conventional model-by-model risk assessment is insufficient and that governance should target system-level interactions and collective-action structures.

Significance. If the framework is accepted, it shifts the focus of AI risk governance from frontier-model capabilities and individual provider liability to the structure of sociotechnical systems: market concentration, value-chain integration, information asymmetries, and institutional incentives. The paper's strengths are its broad and careful synthesis of literatures from complexity science, economics, law, and AI governance; its explicit recognition of its own limitations; and its useful catalog of mechanisms with citations to recent empirical work. It does not provide formal models, new data, or falsifiable predictions, but as a conceptual contribution it provides a valuable analytical heuristic. The central risk is that the strong conclusion—that conventional risk assessment is insufficient—rests on a strong-emergence premise that is asserted rather than demonstrated, and whose scope is left underspecified.

major comments (2)
  1. [Section 3, Section 6.2, Section 8] The paper's central policy conclusion—that 'conventional approaches and methods of risk assessment are insufficient' (Section 8)—depends on the strong-emergence claim that macro-level harms 'cannot be reduced to the features of the individual parts' (Section 3). This claim is justified only by analogies to finance, climate, and social discrimination, not by evidence or an operational criterion for AI-related societal harms. The paper's own examples complicate the claim: algorithmic monoculture and outcome homogenisation (Section 5.6) can be fully explained by correlated application of the same model across many decisions, which is a statistical/aggregative phenomenon rather than irreducible emergence; and Section 6.2 explicitly lists 'accumulation of individual risks' as a source of emergent phenomena, which is additive by definition. The authors should either (a) weaken the conclusion t
  2. [Section 7 vs. Section 8] Section 7 openly acknowledges that 'tipping points... have not yet been sufficiently researched' and that the systemic implications of outcome homogenisation 'have yet to be adequately researched.' These admissions are appropriate and strengthen the paper's credibility, but they are not reflected in the categorical conclusion of Section 8. If the existence and empirical relevance of non-linear amplification, feedback dynamics, and tipping points in AI-induced societal harms are open questions, then the conclusion that conventional risk assessment is 'insufficient' should be framed as an implication of the proposed conceptual lens, not as an established empirical finding. The authors should specify what kind of evidence would be needed to confirm or refute the framework's central premise, or explicitly present the framework as a research agenda rather than a settled conclusion.
minor comments (4)
  1. [Abstract] Line: 'have not sufficiently take complexity' should read 'have not sufficiently taken complexity'; the following sentence needs 'into account' after 'complexity and emergence'.
  2. [Section 5.3] The citation 'Shavit et al.' appears without a year in the text; the reference list gives the year 2023. Please add the year to the in-text citation.
  3. [Section 3] The term 'emergence' is used in a strong, irreducibility sense at first mention. Consider clarifying at that point whether the authors mean ontological or epistemic emergence, since the later argument relies on the irreducibility claim and the distinction matters for what would count as evidence.
  4. [Section 6.2] The phrase 'Emergent phenomena can result from the accumulation of individual risks' is terminologically confusing, since accumulation is not emergence in the strong sense used earlier. Consider using 'macro-level effects' or 'systemic outcomes' instead of 'emergent phenomena' in this passage.

Circularity Check

0 steps flagged

Conceptual framework paper with no fitted inputs, no derived predictions, and no load-bearing self-citations; the policy conclusion is transparently conditional on its own proposed conceptualization, and the paper candidly flags its missing empirical support.

full rationale

This is a conceptual/framework paper, not an empirical or formal derivation. Section 1 explicitly disclaims derivation: the paper mainly provides 'an informed conceptualisation and abstraction from common themes in previous research… neither a complete inventory… nor a critical examination of inconsistencies of concepts, but rather an outline of common systemic risk phenomena and a proposal for an analytical heuristic to be further expanded upon.' The central claim (Section 8) is that systemic AI risks should be conceptualized as emergent macro-level harms arising from collective action problems and externalities, and that consequently conventional model-by-model risk assessment is insufficient. That conclusion is analytically conditional on the paper's own proposed definition of systemic risk as emergent and non-additive — 'Based on this understanding of systemic risks, it can be concluded that conventional approaches and methods of risk assessment are insufficient' — and the paper nowhere claims to have empirically verified that AI harms are non-additive. To the contrary, Section 7 candidly admits that 'the tipping points… have not yet been sufficiently researched' and that systemic implications 'have yet to be adequately researched, particularly in systemic contexts.' There is no fitted parameter renamed as a prediction, no self-definitional equation, no uniqueness theorem imported from the authors' own prior work, and no ansatz smuggled in via citation. The self-citations (Orwat 2024; Orwat et al. 2024; Gazos et al. 2025; Staab et al. 2024) support subsidiary empirical or normative claims — e.g., Staab et al. (2024) on LLM inference is an externally verifiable, peer-reviewed result — and are not load-bearing for the central framework, which rests on external literatures (Mitchell 2009; Ostrom 1990; Renn et al. 2022; Trantidis 2024). The skeptic's concern that the macro-level emergence premise is asserted by analogy rather than tested for AI is a correctness/evidence gap, not circularity: no specific reduction of an output to an input can be exhibited.

Axiom & Free-Parameter Ledger

0 free parameters · 5 axioms · 0 invented entities

The central claim rests on definitions and causal-conceptual premises rather than on parameters or new postulated entities. No free parameters and no invented entities were introduced. The five listed axioms are substantive domain assumptions, stated openly in the text but not empirically established here.

axioms (5)
  • domain assumption Emergent macro-level properties cannot be reduced to micro-level components.
    The entire emergence framing depends on non-reducibility. If AI harms were simply additive aggregates, system-level assessment would reduce to summing individual risks, which the paper explicitly rejects.
  • domain assumption Societal goals—fundamental rights, democracy, sustainability—are emergent properties of sociotechnical systems that AI can threaten at macro level.
    The paper's 'systemic risk' label presupposes that the good being protected is itself a system-level effect rather than a mere collection of individual entitlements.
  • domain assumption The EU AI Act's protection goals (fundamental rights, environment, democracy, rule of law) are the appropriate normative yardstick.
    The paper declares this in Section 1, and its policy conclusions about inadequate governance and the need for societal acceptability judgments depend on adopting that EU value set.
  • domain assumption Collective action problems and complex externalities are the main causal contributions to systemic AI risks.
    This is the paper's core causal claim. It is argued by synthesis of analogous literatures (finance, climate, discrimination) but not demonstrated by data or formal model.
  • domain assumption Market concentration, algorithmic monoculture, feedback loops, information asymmetries, and institutional deficits interact non-linearly to produce systemic harms.
    These amplifiers are imported from complexity science and related fields; their transfer to AI sociotechnical systems is plausible but asserted rather than measured.

pith-pipeline@v1.3.0-alltime-deepseek · 39565 in / 13964 out tokens · 137895 ms · 2026-08-01T15:45:58.963544+00:00 · methodology

0 comments
read the original abstract

The integration of general-purpose artificial intelligence models into downstream AI systems, among other developments, has given rise to new forms of risk that are more systemic in nature than conventional AI risks. However, there is no generally accepted definition of systemic risks in general and for AI in particular. Conceptualisations of these risks vary across research and regulation. Especially the application of the systemic risk approach to human rights or fundamental rights, like in the EU AI Act, is relatively new, just as the research on the contribution of AI to systemic forms of discrimination, privacy violations, erosions of democracy, or climate and environmental degradation. We argue that some concepts so far have not sufficiently take complexity and emergence into account. Furthermore, this variety of concepts might hinder responsible actors to adequately assess the systemic risks of AI, leading to inadequate prevention and mitigation measures and ineffective governance. To contribute to the understanding of systemic risks of AI, we propose a conceptualisation of systemic risks of AI that considers complex phenomena that lead to the emergence of harms at the societal or global level. We outline systemic risks mainly as complex externalities and collective action problems. Of particular interest are feedback dynamics, processes that lead to market concentration like network effects, algorithmic monocultures, and integration processes of AI supply chains or 'AI ecosystems' and across societal sectors, which can result in structural dominance, (inter-) dependencies, and cascading risks. Further phenomena contributing to systemic risks are information asymmetries, informational emergence, and deficits of the governance and institutional framework.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 7 linked inside Pith

  1. [1]

    (2026) Integrating tipping point concepts across diverse systems

    Abrams, J.F., Pereira, L.M., Spaiser, V., Wunderling, N., Villasante, S., Tàbara, J.D., Biggs, R., Armstrong McKay, D.I., Zimm, C., Milkoreit, M., Constantino, S.M., Debnath, R., Dakos, V., Ritchie, P.D.L., Smith, S.R., Rocha, J.C., Kéfi, S., Buxton, J.E., Boulton, C.A., Greenlees, K., Juhola, S., Bhowmik, A., Fesenfeld, L.P., Clarke, J., Dyke, J.G., Layb...

    Pith/arXiv arXiv 2026

  2. [12]

    Hansson, S.O

    Coorporative AI Foundation. Hansson, S.O. (2013) The Ethics of Risk: Ethical Analysis in an Uncertain World. Cham: Palgrave Macmillan. Hardin, G. (1968) The Tragedy of the Commons. Science 162(3859):1243–1248. Hardt, M., Mendler -Dünner, C. (2025) Performative Prediction: Past and Future. Statistical Science 40(3):417–436. Hedström, P., Ylikoski, P. (2010...

    arXiv 2013

  3. [16]

    Centeno, M.A., Nag, M., Patterson, T.S., Shaver, A., Windawi, A.J

    Frankfurt am Main: European Systemic Risk Board. Centeno, M.A., Nag, M., Patterson, T.S., Shaver, A., Windawi, A.J. (2015) The Emergence of Global Systemic Risk. Annual Review of Sociology 41:65–85. 33 Chambers, C., Rade, L., Patterson, T., Centeno, M.A. (2019) The system made me do it? Regulating systemic risk. In: Galaz V. (ed) Global Challenges, Govern...

    Pith/arXiv arXiv 2015

  4. [25]

    Everyone wants to do the model work, not the data work

    Human- AI coevolution. Artificial Intelligence 339:article no. 104244. Perrow, C.B. (1984) Normal Accidents: Living with High-Risk Technologies. New York: Basic Books. Pierson, P. (2000) Increasing returns, path dependence, and the study of politics. American Political Science Review 94(2):251-267. Rathje, S., Van Bavel, J.J., Van Der Linden, S. (2021) Ou...

    1984

  5. [27]

    (2025) The impact of advanced AI systems on democracy

    Summerfield, C., Argyle, L.P., Bakker, M., Collins, T., Durmus, E., Eloundou, T., Gabriel, I., Ganguli, D., Hackenburg, K., Hadfield, G.K., Hewitt, L., Huang, S., Landemore, H., Marchal, N., Ovadya, A., Procaccia, A., Risse, M., Schneier, B., Seger, E., Si ddarth, D., Skaug Sætra, H., Tessler, M.H., Botvinick, M. (2025) The impact of advanced AI systems o...

    arXiv 2025

  6. [32]

    (2022) The limitations of privacy rights

    Solove, D.J. (2022) The limitations of privacy rights. Notre Dame Law Review 98(3):975–1035. Solove, D.J. (2025) Artificial intelligence and privacy. Florida Law Review 77:1–73. Solow-Niederman, A. (2022) Information privacy and the inference economy. Northwestern University Law Review 117:357. Sourati, Z., S. Ziabari, A., Dehghani, M. (2026) The homogeni...

    2022

  7. [36]

    (2024) Government externalities

    Trantidis, A. (2024) Government externalities. Public Choice 201(3):451–469. Trist, E.L. (1981) The evolution of socio-technical systems, No. 2 Ontario Quality of Working Life Centre Toronto. Tsvetkova, M., Yasseri, T., Pescetelli, N., Werner, T. (2024) A new sociology of humans and machines. Nature Human Behaviour 8(10):1864–1876. Tucker, C. (2019) Priva...

    2024

  8. [37]

    Andrews, M., Smart, A., Birhane, A

    Paris: OECD Publishing. Andrews, M., Smart, A., Birhane, A. (2024) The reanimation of pseudoscience in machine learning and its ethical repercussions. Patterns 9(9):article 101027. Arifovic, J., He, X.- z., Wei, L. (2022) Machine learning and speed in high- frequency trading. Journal of Economic Dynamics and Control 139:article 104438. Arnscheidt, C.W., B...

    Pith/arXiv arXiv 2024

  9. [53]

    (2025) The Systemic Environmental Risks of Artificial Intelligence

    Schön, J., Hoffmann, L., Becker, N. (2025) The Systemic Environmental Risks of Artificial Intelligence . Berlin: Gesellschaft für Informatik. Schroeder, D.T., Cha, M., Baronchelli, A., Bostrom, N., Christakis, N.A., Garcia, D., Goldenberg, A., Kyrychenko, Y., Leyton- Brown, K., Lutz, N., Marcus, G., Menczer, F., Pennycook, G., Rand, D.G., Ressa, M., Schwe...

    2025

  10. [85]

    (2025) Concentrating intelligence: scaling and market structure in artificial intelli- gence

    37 Korinek, A., Vipra, J. (2025) Concentrating intelligence: scaling and market structure in artificial intelli- gence. Economic Policy 40(121):225–256. Kosinski, M., Khambatta, P., Wang, Y. (2024) Facial recognition technology and human raters can predict political orientation from images of expressionless faces even when controlling for demographics and...

    2025

  11. [98]

    (2025) The sum of all (un)intentions: reasonable foreseeability, platform algorithms, and emergent systemic harm to marginalized communities

    Khoo, C. (2025) The sum of all (un)intentions: reasonable foreseeability, platform algorithms, and emergent systemic harm to marginalized communities. In: Calo R., Froomkin A.M., Thomasen K. (eds) Robot Law: Volume II. Aldershot, Northampton: Edward Elgar, pp. 106–197. Kianpour, M., Kowalski, S.J., Øverby, H. (2022) Advancing the concept of cybersecurity ...

    2025

  12. [118]

    Automated data analysis

    Dobusch, L., Schüßler, E. (2013) Theorizing path dependence: a review of positive feedback mechanisms in technology markets, regional clusters, and organizations. Industrial and Corporate Change 22(3):617–647. Domínguez Hernández, A., Krishna, S., Perini, A.M., Katell, M., Bennett, S., Borda, A., Hashem, Y., Hadjiloizou, S., Mahomed, S., Jayadeva, S. (202...

    2013

  13. [124]

    Olugbade, O

    Cambridge, Mass.: Harvard University Press. Olugbade, O. (2025) In search of a global governance mechanism for Artificial Intelligence (AI): a collec- tive action perspective. Global Public Policy and Governance 5(2):139–161. Orwat, C. (2020) Risks of Discrimination through the Use of Algorithms. A study compiled with a grant from the Federal Anti-Discrim...

    2025

  14. [138]

    (2025) The complexity of misinformation extends beyond virus and warfare analogies

    Frischlich, L., Olsson, H., Roy, A., Schulze, H., Rhodes, S., Mansheim, A. (2025) The complexity of misinformation extends beyond virus and warfare analogies. npj Complexity 2(1):Article no. 29, pp. 1–8. FTC (2024) A Look Behind the Screens. Examining the Data Practices of Social Media and Video Streaming Services. Washington: Federal Trade Commission (FT...

    Pith/arXiv arXiv 2025

  15. [214]

    (2025) Lessons from complex systems science for AI governance

    Kolt, N., Shur -Ofry, M., Cohen, R. (2025) Lessons from complex systems science for AI governance. Patterns 6(8):article 101341. Kondor, D., Hafez, V., Shankar, S., Wazir, R., Karimi, F. (2024) Complex systems perspective in assessing risks in artificial intelligence. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineerin...

    2025

  16. [289]

    (2024) The Artificial Recruiter: Risks of Discrimination in Employ- ers’ Use of AI and Automated Decision‐Making

    Larsson, S., White, J.M., Bogusz, C.I. (2024) The Artificial Recruiter: Risks of Discrimination in Employ- ers’ Use of AI and Automated Decision‐Making. Social Inclusion 12:article

    2024

  17. [302]

    (2025) Position: Privacy Is Not Just Memorization! ArXiv:2510.01645 Mitchell, M

    Mireshghallah, N., Li, T. (2025) Position: Privacy Is Not Just Memorization! ArXiv:2510.01645 Mitchell, M. (2009) Complexity: A Guided Tour. Oxford: Oxford University Press. Mittelstadt, B. (2017) From Individual to Group Privacy in Big Data Analytics. Philosophy & Technology 30:475–494. Mittelstadt, B.D., Allo, P., Taddeo, M., Wachter, S., Floridi, L. (2...

    arXiv 2025

  18. [313]

    (2024) Regulate against the machine: how the EU mitigates AI harm to democracy

    Cupać, J., Sienknecht, M. (2024) Regulate against the machine: how the EU mitigates AI harm to democracy. Democratization 31(5):1067–1090. Darius, P., Hoppe, T., Aleksandrov, A. (2025) Systemic Risks of Interacting AI. Berlin: Fraunhofer-Institut für Offene Kommunikationssysteme. David, P.A. (1985) Clio and the Economics of QWERTY. American Economic Revie...

    2024

  19. [352]

    Gur-Arieh, S., Lee, C. (2025) Consistently Arbitrary or Arbitrarily Consistent: Navigating the Tensions Between Homogenization and Multiplicity in Algorithmic Decision -Making, at: Proceedings of the 2025 ACM Conference on Fairness, Accountability, and Transparency. Hacker, P., Kasirzadeh, A., Edwards, L. (2025) AI, digital platforms, and the new systemic...

    2025

  20. [437]

    (2021) Public good, collective action and financial regulation

    Ülgen, F. (2021) Public good, collective action and financial regulation. Annals of Public and Cooperative Economics 92(1):147–167. Uuk, R., Gutierrez, C.I., Guppy, D., Lauwaert, L., Kasirzadeh, A., Velasco, L., Slattery, P., Prunkl, C. (2024) A Taxonomy of Systemic Risks from General -Purpose AI. arXiv preprint arXiv:2412.07780. van Eeten, M.J.G., Bauer,...

    Pith/arXiv arXiv 2021

  21. [1563]

    (2016) The foundations of the environmental rebound effect and its contribution towards a general framework

    Vivanco, D.F., McDowall, W., Freire-González, J., Kemp, R., van der Voet, E. (2016) The foundations of the environmental rebound effect and its contribution towards a general framework. Ecological economics 125:60–69. Vlek, C., Steg, L. (2007) Human Behavior and Environmental Sustainability: Problems, Driving Forces, and Research Topics. Journal of Social...

    Pith/arXiv arXiv 2016

  22. [1912]

    (2012) Network Effects and Social Inequality

    DiMaggio, P., Garip, F. (2012) Network Effects and Social Inequality. Annual Review of Sociology 38:93–

    2012

  23. [2023]

    1 BvR 1547/19

    Docket No. 1 BvR 1547/19. Official English translation by the Court. ECLI:DE:BVerfG:2023:rs20230216.1bvr154719. Karlsruhe: Federal Constitutional Court. Felice, W.F., Fuguitt, D. (2021) Human Rights and Public Goods: The Global New Deal. 3rd ed. Lanham: Rowman & Littlefield. Ferrara, E. (2024) The Butterfly Effect in artificial intelligence systems: Impli...

    2023

  24. [2024]

    (2025) The Backfiring Effect of Weak AI Safety Regulation

    Laufer, B., Kleinberg, J., Heidari, H. (2025) The Backfiring Effect of Weak AI Safety Regulation. arXiv preprint arXiv:2503.20848. Lawrence, M., Homer -Dixon, T., Janzwood, S., Rockstöm, J., Renn, O., Donges, J.F. (2024) Global polycrisis: the causal mechanisms of crisis entanglement. Global Sustainability 7:article no. e6. Lee, H.-P., Yang, Y.-J., Von Da...

    Pith/arXiv arXiv 2025

  25. [2026]

    London: Government UK, Department for Science, Innovation and Technology and AI Safety Institute. Bengio, Y., Privitera, D., Besiroglu, T., Bommasani, R., Casper, S., Choi, Y., Goldfarb, D., Heidari, H., Khalatbari, L., Longpre, S., Mavroudis, V., Mazeika, M., Ng, K.Y., Okolo, C.T., Raji, D., Skeadas, T., Tramèr, F., Mindermann, S. (2025) International AI...

    2025

  26. [2050]

    Page, S.E

    Proceedings of the National Academy of Sciences 117(5):2354–2365. Page, S.E. (2015) What Sociologists Should Know About Complexity. Annual Review of Sociology 41(1):21–41. Palumbo, A. (2026) Charting systemic risk management as a regulatory paradigm in EU digital legislation: features, challenges and directions for future research. Technology and Regulati...

    2015

  27. [4692]

    (2002) Common Goods and Governance

    Mayntz, R. (2002) Common Goods and Governance. In: Héritier A. (ed) Common Goods: Reinventing European and International Governance . Lanham et al.: Rowman & Littlefield Publishers, pp. 15–27. Mayntz, R. (2008) Emergence in Philosophy and Social Theory. In: Mayntz R. (ed) Sozialwissen schaft- liches Erklären. Probleme der Theoriebildung und Methodologie. ...

    2002

  28. [7160]

    (2024) Normative Challenges of Risk Regula - tion of Artificial Intelligence

    Orwat, C., Bareis, J., Folberth, A., Jahnel, J., Wadephul, C. (2024) Normative Challenges of Risk Regula - tion of Artificial Intelligence. NanoEthics 18(2 (Article 11)):1–29. Ostrom, E. (1990) Governing the Commons: the Evolution of Institutions for Collective Action, The Political Economy of Institutions and Decisions. Cambridge, New York: Cambridge Uni...

    2024

  29. [7382]

    (2010) Engaging rational discrimination: exploring reasons for placing regulatory constraints on decision support systems

    Gandy Jr., O.H. (2010) Engaging rational discrimination: exploring reasons for placing regulatory constraints on decision support systems. Ethics and Information Technology 12(1):1–14. Gauthier, G., Hodler, R., Widmer, P., Zhuravskaya, E. (2026) The political effects of X’s feed algorithm. Nature 652:416–423. Gazos, A., Kahn, J., Kusche, I., Büscher, C., ...

    2010

  30. [7471]

    (2024) Fine -Tuning Games: Bargaining and Adaptation for General - Purpose Models, at: WWW '24: Proceedings of the ACM Web Conference

    Laufer, B., Kleinberg, J., Heidari, H. (2024) Fine -Tuning Games: Bargaining and Adaptation for General - Purpose Models, at: WWW '24: Proceedings of the ACM Web Conference

    2024