-
Do LLMs trust AI regulation? Emerging behaviour of game-theoretic LLM agents
Authors:
Alessio Buscemi,
Daniele Proverbio,
Paolo Bova,
Nataliya Balabanova,
Adeela Bashir,
Theodor Cimpeanu,
Henrique Correia da Fonseca,
Manh Hong Duong,
Elias Fernandez Domingos,
Antonio M. Fernandes,
Marcus Krellner,
Ndidi Bianca Ogbo,
Simon T. Powers,
Fernando P. Santos,
Zia Ush Shamszaman,
Zhao Song,
Alessandro Di Stefano,
The Anh Han
Abstract:
There is general agreement that fostering trust and cooperation within the AI development ecosystem is essential to promote the adoption of trustworthy AI systems. By embedding Large Language Model (LLM) agents within an evolutionary game-theoretic framework, this paper investigates the complex interplay between AI developers, regulators and users, modelling their strategic choices under different…
▽ More
There is general agreement that fostering trust and cooperation within the AI development ecosystem is essential to promote the adoption of trustworthy AI systems. By embedding Large Language Model (LLM) agents within an evolutionary game-theoretic framework, this paper investigates the complex interplay between AI developers, regulators and users, modelling their strategic choices under different regulatory scenarios. Evolutionary game theory (EGT) is used to quantitatively model the dilemmas faced by each actor, and LLMs provide additional degrees of complexity and nuances and enable repeated games and incorporation of personality traits. Our research identifies emerging behaviours of strategic AI agents, which tend to adopt more "pessimistic" (not trusting and defective) stances than pure game-theoretic agents. We observe that, in case of full trust by users, incentives are effective to promote effective regulation; however, conditional trust may deteriorate the "social pact". Establishing a virtuous feedback between users' trust and regulators' reputation thus appears to be key to nudge developers towards creating safe AI. However, the level at which this trust emerges may depend on the specific LLM used for testing. Our results thus provide guidance for AI regulation systems, and help predict the outcome of strategic LLM agents, should they be used to aid regulation itself.
△ Less
Submitted 11 April, 2025;
originally announced April 2025.
-
Resilience of the autocatalytic feedback loop for gene regulation
Authors:
Daniele Proverbio,
Giulia Giordano
Abstract:
Gene expression in response to stimuli is regulated by transcription factors (TFs) through feedback loop motifs, aimed at maintaining the desired TF concentration despite uncertainties and perturbations. In this work, we consider a stochastic model of the autocatalytic feedback loop for gene regulation and we probabilistically quantify its resilience, \textit{i.e.}, its ability to preserve the equ…
▽ More
Gene expression in response to stimuli is regulated by transcription factors (TFs) through feedback loop motifs, aimed at maintaining the desired TF concentration despite uncertainties and perturbations. In this work, we consider a stochastic model of the autocatalytic feedback loop for gene regulation and we probabilistically quantify its resilience, \textit{i.e.}, its ability to preserve the equilibrium associated with a prescribed concentration of TFs, and the corresponding basin of attraction, in the presence of noise. We show that the formation of larger oligomers, corresponding to larger Hill coefficients of the regulation function, and thus to sharper non-linearities, improves the system resilience, even close to critical concentrations of TFs. Our formal results within a stochastic formulation relying on the Fokker-Planck equation are accompanied by numerical simulations.
△ Less
Submitted 4 April, 2025;
originally announced April 2025.
-
Media and responsible AI governance: a game-theoretic and LLM analysis
Authors:
Nataliya Balabanova,
Adeela Bashir,
Paolo Bova,
Alessio Buscemi,
Theodor Cimpeanu,
Henrique Correia da Fonseca,
Alessandro Di Stefano,
Manh Hong Duong,
Elias Fernandez Domingos,
Antonio Fernandes,
The Anh Han,
Marcus Krellner,
Ndidi Bianca Ogbo,
Simon T. Powers,
Daniele Proverbio,
Fernando P. Santos,
Zia Ush Shamszaman,
Zhao Song
Abstract:
This paper investigates the complex interplay between AI developers, regulators, users, and the media in fostering trustworthy AI systems. Using evolutionary game theory and large language models (LLMs), we model the strategic interactions among these actors under different regulatory regimes. The research explores two key mechanisms for achieving responsible governance, safe AI development and ad…
▽ More
This paper investigates the complex interplay between AI developers, regulators, users, and the media in fostering trustworthy AI systems. Using evolutionary game theory and large language models (LLMs), we model the strategic interactions among these actors under different regulatory regimes. The research explores two key mechanisms for achieving responsible governance, safe AI development and adoption of safe AI: incentivising effective regulation through media reporting, and conditioning user trust on commentariats' recommendation. The findings highlight the crucial role of the media in providing information to users, potentially acting as a form of "soft" regulation by investigating developers or regulators, as a substitute to institutional AI regulation (which is still absent in many regions). Both game-theoretic analysis and LLM-based simulations reveal conditions under which effective regulation and trustworthy AI development emerge, emphasising the importance of considering the influence of different regulatory regimes from an evolutionary game-theoretic perspective. The study concludes that effective governance requires managing incentives and costs for high quality commentaries.
△ Less
Submitted 12 March, 2025;
originally announced March 2025.
-
Functional observability and subspace reconstruction in nonlinear systems
Authors:
Arthur N. Montanari,
Leandro Freitas,
Daniele Proverbio,
Jorge Gonçalves
Abstract:
Time-series analysis is fundamental for modeling and predicting dynamical behaviors from time-ordered data, with applications in many disciplines such as physics, biology, finance, and engineering. Measured time-series data, however, are often low dimensional or even univariate, thus requiring embedding methods to reconstruct the original system's state space. The observability of a system establi…
▽ More
Time-series analysis is fundamental for modeling and predicting dynamical behaviors from time-ordered data, with applications in many disciplines such as physics, biology, finance, and engineering. Measured time-series data, however, are often low dimensional or even univariate, thus requiring embedding methods to reconstruct the original system's state space. The observability of a system establishes fundamental conditions under which such reconstruction is possible. However, complete observability is too restrictive in applications where reconstructing the entire state space is not necessary and only a specific subspace is relevant. Here, we establish the theoretic condition to reconstruct a nonlinear functional of state variables from measurement processes, generalizing the concept of functional observability to nonlinear systems. When the functional observability condition holds, we show how to construct a map from the embedding space to the desired functional of state variables, characterizing the quality of such reconstruction. The theoretical results are then illustrated numerically using chaotic systems with contrasting observability properties. By exploring the presence of functionally unobservable regions in embedded attractors, we also apply our theory for the early warning of seizure-like events in simulated and empirical data. The studies demonstrate that the proposed functional observability condition can be assessed a priori to guide time-series analysis and experimental design for the dynamical characterization of complex systems.
△ Less
Submitted 10 January, 2023;
originally announced January 2023.
-
Buffering variability in cell regulation motifs close to criticality
Authors:
Daniele Proverbio,
Arthur N. Montanari,
Alexander Skupin,
Jorge Gonçalves
Abstract:
Bistable biological regulatory systems need to cope with stochastic noise to fine-tune their function close to bifurcation points. Here, we study stability properties of this regime in generic systems to demonstrate that cooperative interactions buffer system variability, hampering noise-induced regime shifts. Our analysis also shows that, in the considered cooperativity range, impending regime sh…
▽ More
Bistable biological regulatory systems need to cope with stochastic noise to fine-tune their function close to bifurcation points. Here, we study stability properties of this regime in generic systems to demonstrate that cooperative interactions buffer system variability, hampering noise-induced regime shifts. Our analysis also shows that, in the considered cooperativity range, impending regime shifts can be generically detected by statistical early warning signals from distributional data. Our generic framework, based on minimal models, can be used to extract robustness and variability properties of more complex models and empirical data close to criticality.
△ Less
Submitted 16 December, 2022;
originally announced December 2022.
-
Robust synchronization of heterogeneous robot swarms on the sphere
Authors:
Johan Markdahl,
Daniele Proverbio,
Jorge Goncalves
Abstract:
Synchronization on the sphere is important to certain control applications in swarm robotics. Of recent interest is the Lohe model, which generalizes the Kuramoto model from the circle to the sphere. The Lohe model is mainly studied in mathematical physics as a toy model of quantum synchronization. The model makes few assumptions, wherefore it is well-suited to represent a swarm. Previous work on…
▽ More
Synchronization on the sphere is important to certain control applications in swarm robotics. Of recent interest is the Lohe model, which generalizes the Kuramoto model from the circle to the sphere. The Lohe model is mainly studied in mathematical physics as a toy model of quantum synchronization. The model makes few assumptions, wherefore it is well-suited to represent a swarm. Previous work on this model has focused on the cases of complete and acyclic networks or the homogeneous case where all oscillator frequencies are equal. This paper concerns the case of heterogeneous oscillators connected by a non-trivial network. We show that any undesired equilibrium is exponentially unstable if the frequencies satisfy a given bound. This property can also be interpreted as a robustness result for small model perturbations of the homogeneous case with zero frequencies. As such, the Lohe model is a good choice for control applications in swarm robotics.
△ Less
Submitted 1 April, 2020;
originally announced April 2020.
-
Dynamical strategies for obstacle avoidance during Dictyostelium discoideum aggregation: a Multi-agent system model
Authors:
Daniele Proverbio,
Marco Maggiora
Abstract:
Chemotaxis, the movement of an organism in response to chemical stimuli, is a typical feature of many microbiological systems. In particular, the social amoeba \textit{Disctyostelium discoideum} is widely used as a model organism, but it is not still clear how it behaves in heterogeneous environments. A few models focusing on mechanical features have already addressed the question; however, we sug…
▽ More
Chemotaxis, the movement of an organism in response to chemical stimuli, is a typical feature of many microbiological systems. In particular, the social amoeba \textit{Disctyostelium discoideum} is widely used as a model organism, but it is not still clear how it behaves in heterogeneous environments. A few models focusing on mechanical features have already addressed the question; however, we suggest that phenomenological models focusing on the population dynamics may provide new meaningful data. Consequently, by means of a specific Multi-agent system model, we study the dynamical features emerging from complex social interactions among individuals belonging to amoeba colonies.\\ After defining an appropriate metric to quantitatively estimate the gathering process, we find that: a) obstacles play the role of local topological perturbation, as they alter the flux of chemical signals; b) physical obstacles (blocking the cellular motion and the chemical flux) and purely chemical obstacles (only interfering with chemical flux) elicit similar dynamical behaviors; c) a minimal program for robustly gathering simulated cells does not involve mechanisms for obstacle sensing and avoidance; d) fluctuations of the dynamics concur in preventing multiple stable clusters. Comparing those findings with previous results, we speculate about the fact that chemotactic cells can avoid obstacles by simply following the altered chemical gradient. Social interactions are sufficient to guarantee the aggregation of the whole colony past numerous obstacles.
△ Less
Submitted 16 December, 2019; v1 submitted 13 May, 2019;
originally announced May 2019.
-
A Multi-agent approach for $\textit{in silico}$ simulations of micro-biological systems
Authors:
Daniele Proverbio,
Luca Gallo,
Barbara Passalacqua,
Jacopo Pellegrino,
Marco Maggiora
Abstract:
Using a Multi-agent systems paradigm, the present project develops, validates and exploits a computational $\textit{testbed}$ that simulates micro-biological complex systems, namely the aggregation patterns of the social amoeba $\textit{Dyctiostelium discoideum}$. We propose a new design and implementation for managing discrete simulations with autonomous agents on a microscopic scale, thus focusi…
▽ More
Using a Multi-agent systems paradigm, the present project develops, validates and exploits a computational $\textit{testbed}$ that simulates micro-biological complex systems, namely the aggregation patterns of the social amoeba $\textit{Dyctiostelium discoideum}$. We propose a new design and implementation for managing discrete simulations with autonomous agents on a microscopic scale, thus focusing on their social behavior and mutual interactions. Then, the dependence on the main physical variables is tested, namely density and number of amoebas; in addition, we analyze the robustness of the dynamics against various noise sources. Along with these results, we suggest a methodology for further studies that make use of our validated model.
△ Less
Submitted 16 April, 2019;
originally announced April 2019.
