Konstantinos Voudouris
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Konstantinos Voudouris has a total of 103 co-authors. Here are the top co-authors by number of shared publications:
Eric Schulz
Institute for Human-Centered AI, Helmholtz Munich, Neuherberg, Germany
3 shared publications- Lucy Gaia Cheke
Department of Psychology, University of Cambridge, Cambridge, United Kingdom
3 shared publications Marta Halina
Leverhulme Centre for the Future of Intelligence, University of Cambridge
3 shared publications- Benjamin George Farrar
Department of Psychology, University of Cambridge
2 shared publications Elif Akata
2 shared publications- John Burden
Centre for the Study of Existential Risk
2 shared publications
Latest Publications
- Open Access
- 29 October 2026
- Published by Springer Nature in Minds and Machines
- Vol. 36 (4) , 50
Measuring What AI Systems Might Do: Towards A Measurement Science in AI
Scientists, policy-makers, business leaders, and members of the public care about what modern artificial intelligence systems are disposed to do. Yet terms such as capabilities , propensities , skills , values , and abilities are routinely used interchangeably and conflated with observable performance, with AI evaluation practices rarely specifying what quantity they purport to measure. Building on established dispositional accounts of ability, we argue that capabilities and propensities are dispositional properties —stable features of systems characterised by counterfactual relationships between contextual conditions and behavioural outputs. Measuring a disposition requires (i) hypothesising which contextual properties are causally relevant, (ii) independently operationalising and measuring those properties, and (iii) empirically mapping how variation in those properties affects the probability of the behaviour. Benchmark averages and data-driven latent-variable models like Item Response Theory can provide useful evidence, but do not by themselves identify these causal structures. By developing this principled account of capabilities, propensities, and how to measure them, we are able to outline what disposition-respecting, scientifically defensible AI evaluation would require.
- 29 October 2026
- Published by Springer Nature in Minds and Machines
- Vol. 36 (4) , 50
Open AccessMeasuring What AI Systems Might Do: Towards A Measurement Science in AI
Scientists, policy-makers, business leaders, and members of the public care about what modern artificial intelligence systems are disposed to do. Yet terms such as capabilities , propensities , skills , values , and abilities are routinely used interchangeably and conflated with observable performance, with AI evaluation practices rarely specifying what quantity they purport to measure. Building on established dispositional accounts of ability, we argue that capabilities and propensities are dispositional properties —stable features of systems characterised by counterfactual relationships between contextual conditions and behavioural outputs. Measuring a disposition requires (i) hypothesising which contextual properties are causally relevant, (ii) independently operationalising and measuring those properties, and (iii) empirically mapping how variation in those properties affects the probability of the behaviour. Benchmark averages and data-driven latent-variable models like Item Response Theory can provide useful evidence, but do not by themselves identify these causal structures. By developing this principled account of capabilities, propensities, and how to measure them, we are able to outline what disposition-respecting, scientifically defensible AI evaluation would require.
- Open Access
- 17 July 2026
- Published by Elsevier in SSRN Electronic Journal
Judge Hacking in Recursive Debate Protocols: A Call for Solutions
The debate protocol is a promising solution to the problem of supervising systems that are more capable than their supervisors, known as scalable oversight. It is proposed not only as an inferencetime procedure for adjudicating individual claims, but as a training protocol: the judge's verdict propagates back through the debate as a reward signal used to optimise the debaters. The judge here can be a human or some proxy for human feedback. Under idealised assumptions, debate incentivises truthfulness: the truthful claim is the uniquely optimal one to defend. Real judges, however, make systematic errors. Human judges exhibit predictable cognitive and domain-specific biases, and AI proxies exhibit predictable hallucinations and distributional failures. We argue that when these leaf-level errors are systematic and learnable, a dishonest debater can sometimes win debates while defending a falsehood or attacking truth by strategically steering the debate towards the judge's blind spots. We call this judge hacking. We ground this intuition in a formal game-theoretic model over recursive debate decomposition trees, where judge hacking occurs when the value of a false claim exceeds that of the corresponding true claim, and we speculate that the same mechanism plausibly arises across other variants of debate such as cross-examination and prover-estimator. The risk is most acute when debate is used for training because it introduces strong optimisation pressure to exploit the judge when it is valuable to do so. If debate is to be a viable oversight strategy, judge hacking must be mitigated. We study juries-panels of multiple judges-as a general, judge-side mitigation in need of empirical research. Independent, diverse juries can suppress idiosyncratic errors through a wisdom-of-the-crowd effect, but correlated or homogeneous juries can amplify shared systematic biases. We identify the maximum shared false-positive and false-negative bias over adversarially reachable leaves, together with the correlation structure of juror errors, as the key parameters governing jury robustness. We conclude with a call for urgent empirical and theoretical work on jury design for debate.
- 17 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessJudge Hacking in Recursive Debate Protocols: A Call for Solutions
The debate protocol is a promising solution to the problem of supervising systems that are more capable than their supervisors, known as scalable oversight. It is proposed not only as an inferencetime procedure for adjudicating individual claims, but as a training protocol: the judge's verdict propagates back through the debate as a reward signal used to optimise the debaters. The judge here can be a human or some proxy for human feedback. Under idealised assumptions, debate incentivises truthfulness: the truthful claim is the uniquely optimal one to defend. Real judges, however, make systematic errors. Human judges exhibit predictable cognitive and domain-specific biases, and AI proxies exhibit predictable hallucinations and distributional failures. We argue that when these leaf-level errors are systematic and learnable, a dishonest debater can sometimes win debates while defending a falsehood or attacking truth by strategically steering the debate towards the judge's blind spots. We call this judge hacking. We ground this intuition in a formal game-theoretic model over recursive debate decomposition trees, where judge hacking occurs when the value of a false claim exceeds that of the corresponding true claim, and we speculate that the same mechanism plausibly arises across other variants of debate such as cross-examination and prover-estimator. The risk is most acute when debate is used for training because it introduces strong optimisation pressure to exploit the judge when it is valuable to do so. If debate is to be a viable oversight strategy, judge hacking must be mitigated. We study juries-panels of multiple judges-as a general, judge-side mitigation in need of empirical research. Independent, diverse juries can suppress idiosyncratic errors through a wisdom-of-the-crowd effect, but correlated or homogeneous juries can amplify shared systematic biases. We identify the maximum shared false-positive and false-negative bias over adversarially reachable leaves, together with the correlation structure of juror errors, as the key parameters governing jury robustness. We conclude with a call for urgent empirical and theoretical work on jury design for debate.
- Open Access
- 30 May 2026
- Published by Center for Open Science in PsyArXiv
AI alignment is a human problem
The challenge of ensuring that artificial intelligence (AI) systems behave in ways that humans prefer is known as the alignment problem. This problem is urgent because we are increasingly delegating consequential tasks to AI agents. The pace, scale, and opacity with which modern AI systems act makes supervising them a formidable challenge: they can solve expert-level problems but exhibit unintuitive failure modes, and remain unconstrained by the social norms and institutional structures that make delegation to human agents tolerably safe. We argue that AI alignment is, largely, a human problem - and that tools from the cognitive and social sciences have a major part to play in solving it. Human judgements are a critical component at every stage of the modern alignment training pipeline, from curating data and learning from our preferences, to auditing AI systems and overseeing them while they complete tasks for us. However, the cognitive and social properties of these judgements remain strikingly under-examined. The central challenge for alignment is to amplify the human supervisory signals on which alignment training depends, so that they remain reliable as AI systems produce ever more behaviour to monitor and increasingly operate beyond unaided human expertise. We argue that this requires a research programme focused on five bottlenecks in human supervision: biased and context-sensitive judgement, plural and contested values, limited attention and throughput, limited counterfactual reasoning, and expertise-limited verification.
- 30 May 2026
- Published by Center for Open Science in PsyArXiv
Open AccessAI alignment is a human problem
The challenge of ensuring that artificial intelligence (AI) systems behave in ways that humans prefer is known as the alignment problem. This problem is urgent because we are increasingly delegating consequential tasks to AI agents. The pace, scale, and opacity with which modern AI systems act makes supervising them a formidable challenge: they can solve expert-level problems but exhibit unintuitive failure modes, and remain unconstrained by the social norms and institutional structures that make delegation to human agents tolerably safe. We argue that AI alignment is, largely, a human problem - and that tools from the cognitive and social sciences have a major part to play in solving it. Human judgements are a critical component at every stage of the modern alignment training pipeline, from curating data and learning from our preferences, to auditing AI systems and overseeing them while they complete tasks for us. However, the cognitive and social properties of these judgements remain strikingly under-examined. The central challenge for alignment is to amplify the human supervisory signals on which alignment training depends, so that they remain reliable as AI systems produce ever more behaviour to monitor and increasingly operate beyond unaided human expertise. We argue that this requires a research programme focused on five bottlenecks in human supervision: biased and context-sensitive judgement, plural and contested values, limited attention and throughput, limited counterfactual reasoning, and expertise-limited verification.
- Open Access
- 30 January 2026
- Published by Elsevier in Artificial Intelligence
- Vol. 353
Predictable artificial intelligence
- 30 January 2026
- Published by Elsevier in Artificial Intelligence
- Vol. 353
Open AccessPredictable artificial intelligence
- Open Access
- 1 November 2025
- Published by American Psychological Association (APA) in Journal of Comparative Psychology
- Vol. 139 (4) , 242-259
Morgan’s canon and the associative–cognitive distinction today: A survey of practitioners.
Philosophers of science and mind have paid increasing attention to the field of comparative psychology. Two recurring points of discussion in the literature are the methodological value of Morgan's Canon and the distinction between associative and cognitive processes. Although the existing literature regularly makes claims about the beliefs and attitudes of comparative psychologists, there are few empirical studies verifying these claims. This article fills this gap by presenting and analyzing the views of over 200 comparative psychologists on Morgan's Canon and the associative-cognitive distinction. We found that while there is some agreement between the claims in the existing literature and the views of the practitioners surveyed here (e.g., that Morgan's Canon is a parsimony principle), there are also surprising points of divergence (e.g., practitioners do not view the associative-cognitive distinction as a version of Morgan's Canon). We intend for this study to inform existing philosophical and theoretical work on perennial questions regarding how to study animal minds and behavior. (PsycInfo Database Record (c) 2025 APA, all rights reserved).
- 1 November 2025
- Published by American Psychological Association (APA) in Journal of Comparative Psychology
- Vol. 139 (4) , 242-259
Open AccessMorgan’s canon and the associative–cognitive distinction today: A survey of practitioners.
Philosophers of science and mind have paid increasing attention to the field of comparative psychology. Two recurring points of discussion in the literature are the methodological value of Morgan's Canon and the distinction between associative and cognitive processes. Although the existing literature regularly makes claims about the beliefs and attitudes of comparative psychologists, there are few empirical studies verifying these claims. This article fills this gap by presenting and analyzing the views of over 200 comparative psychologists on Morgan's Canon and the associative-cognitive distinction. We found that while there is some agreement between the claims in the existing literature and the views of the practitioners surveyed here (e.g., that Morgan's Canon is a parsimony principle), there are also surprising points of divergence (e.g., practitioners do not view the associative-cognitive distinction as a version of Morgan's Canon). We intend for this study to inform existing philosophical and theoretical work on perennial questions regarding how to study animal minds and behavior. (PsycInfo Database Record (c) 2025 APA, all rights reserved).
- Open Access
Exploring Major Transitions in the Evolution of Biological Cognition With Artificial Neural Networks
Transitional accounts of evolution emphasise a few changes that shape what is evolvable, with dramatic consequences for derived lineages. More recently it has been proposed that cognition might also have evolved via a series of major transitions that manipulate the structure of biological neural networks, fundamentally changing the flow of information. We used idealised models of information flow, artificial neural networks (ANNs), to evaluate whether changes in information flow in a network can yield a transitional change in cognitive performance. We compared networks with feed-forward, recurrent and laminated topologies, and tested their performance learning artificial grammars that differed in complexity, controlling for network size and resources. We documented a qualitative expansion in the types of input that recurrent networks can process compared to feed-forward networks, and a related qualitative increase in performance for learning the most complex grammars. We also noted how the difficulty in training recurrent networks poses a form of transition barrier and contingent irreversibility -- other key features of evolutionary transitions. Not all changes in network topology confer a performance advantage in this task set. Laminated networks did not outperform non-laminated networks in grammar learning. Overall, our findings show how some changes in information flow can yield transitions in cognitive performance.
Open AccessExploring Major Transitions in the Evolution of Biological Cognition With Artificial Neural Networks
Transitional accounts of evolution emphasise a few changes that shape what is evolvable, with dramatic consequences for derived lineages. More recently it has been proposed that cognition might also have evolved via a series of major transitions that manipulate the structure of biological neural networks, fundamentally changing the flow of information. We used idealised models of information flow, artificial neural networks (ANNs), to evaluate whether changes in information flow in a network can yield a transitional change in cognitive performance. We compared networks with feed-forward, recurrent and laminated topologies, and tested their performance learning artificial grammars that differed in complexity, controlling for network size and resources. We documented a qualitative expansion in the types of input that recurrent networks can process compared to feed-forward networks, and a related qualitative increase in performance for learning the most complex grammars. We also noted how the difficulty in training recurrent networks poses a form of transition barrier and contingent irreversibility -- other key features of evolutionary transitions. Not all changes in network topology confer a performance advantage in this task set. Laminated networks did not outperform non-laminated networks in grammar learning. Overall, our findings show how some changes in information flow can yield transitions in cognitive performance.
- 2 September 2025
- Published by Springer Nature in Biology & Philosophy
- Vol. 40 (5) , 1-20
Analogies and the associative-cognitive distinction in comparative psychology
- 2 September 2025
- Published by Springer Nature in Biology & Philosophy
- Vol. 40 (5) , 1-20
Analogies and the associative-cognitive distinction in comparative psychology
- 13 May 2025
- Published by Springer Nature in Nature Reviews Psychology
- Vol. 4 (6) , 363-364
Bringing comparative cognition approaches to AI systems
- 13 May 2025
- Published by Springer Nature in Nature Reviews Psychology
- Vol. 4 (6) , 363-364
Bringing comparative cognition approaches to AI systems
- Open Access
- 28 February 2025
- Published by Springer Nature in Behavior Research Methods
- Vol. 57 (4) , 1-23
The Animal-AI Environment: A virtual laboratory for comparative cognition and artificial intelligence research
The Animal-AI Environment is a unique game-based research platform designed to facilitate collaboration between the artificial intelligence and comparative cognition research communities. In this paper, we present the latest version of the Animal-AI Environment, outlining several major features that make the game more engaging for humans and more complex for AI systems. These features include interactive buttons, reward dispensers, and player notifications, as well as an overhaul of the environment’s graphics and processing for significant improvements in agent training time and quality of the human player experience. We provide detailed guidance on how to build computational and behavioural experiments with the Animal-AI Environment. We present results from a series of agents, including the state-of-the-art deep reinforcement learning agent Dreamer-v3, on newly designed tests and the Animal-AI testbed of 900 tasks inspired by research in the field of comparative cognition. The Animal-AI Environment offers a new approach for modelling cognition in humans and non-human animals, and for building biologically inspired artificial intelligence.
- 28 February 2025
- Published by Springer Nature in Behavior Research Methods
- Vol. 57 (4) , 1-23
Open AccessThe Animal-AI Environment: A virtual laboratory for comparative cognition and artificial intelligence research
The Animal-AI Environment is a unique game-based research platform designed to facilitate collaboration between the artificial intelligence and comparative cognition research communities. In this paper, we present the latest version of the Animal-AI Environment, outlining several major features that make the game more engaging for humans and more complex for AI systems. These features include interactive buttons, reward dispensers, and player notifications, as well as an overhaul of the environment’s graphics and processing for significant improvements in agent training time and quality of the human player experience. We provide detailed guidance on how to build computational and behavioural experiments with the Animal-AI Environment. We present results from a series of agents, including the state-of-the-art deep reinforcement learning agent Dreamer-v3, on newly designed tests and the Animal-AI testbed of 900 tasks inspired by research in the field of comparative cognition. The Animal-AI Environment offers a new approach for modelling cognition in humans and non-human animals, and for building biologically inspired artificial intelligence.
- Open Access
- 8 November 2024
- Published by Leibniz Institute for Psychology (ZPID) in Biolinguistics
- Vol. 18, 1-15
Metatheoretical linguistics: A philosopher’s guide
In this article, we summarise and critically evaluate Ryan Nefdt’s The Philosophy of Theoretical Linguistics: A Contemporary Outlook (2024). In this book, Nefdt brings the tools of philosophy to bear on contemporary linguistics, targeting perennial debates in syntax, semantics, pragmatics, phonology, and the evolution of language. In so doing, Nefdt sketches several tantalising paths for progress on these topics. Although some of Nefdt’s arguments are underdeveloped, this book serves as a worthwhile introduction to the philosophy of linguistics.
- 8 November 2024
- Published by Leibniz Institute for Psychology (ZPID) in Biolinguistics
- Vol. 18, 1-15
Open AccessMetatheoretical linguistics: A philosopher’s guide
In this article, we summarise and critically evaluate Ryan Nefdt’s The Philosophy of Theoretical Linguistics: A Contemporary Outlook (2024). In this book, Nefdt brings the tools of philosophy to bear on contemporary linguistics, targeting perennial debates in syntax, semantics, pragmatics, phonology, and the evolution of language. In so doing, Nefdt sketches several tantalising paths for progress on these topics. Although some of Nefdt’s arguments are underdeveloped, this book serves as a worthwhile introduction to the philosophy of linguistics.