Unlearnable Games and "Satisficing'' Decisions: A Simple Model for a Complex World

Abstract

As a schematic model of the complexity economic agents are confronted with, we introduce the ``SK-game'', a discrete time binary choice model inspired from mean-field spin-glasses. We show that even in a completely static environment, agents are unable to learn collectively-optimal strategies. This is either because the learning process gets trapped in a sub-optimal fixed point, or because learning never converges and leads to a never ending evolution of agents intentions. Contrarily to the hope that learning might save the standard ``rational expectation'' framework in economics, we argue that complex situations are generically unlearnable and agents must do with satisficing solutions, as argued long ago by Herbert Simon (Simon 1955). Only a centralized, omniscient agent endowed with enormous computing power could qualify to determine the optimal strategy of all agents. Using a mix of analytical arguments and numerical simulations, we find that (i) long memory of past rewards is beneficial to learning whereas over-reaction to recent past is detrimental and leads to cycles or chaos; (ii) increased competition destabilizes fixed points and leads first to chaos and, in the high competition limit, to quasi-cycles; (iii) some amount of randomness in the learning process, perhaps paradoxically, allows the system to reach better collective decisions; (iv) non-stationary, ``aging'' behaviour spontaneously emerge in a large swath of parameter space of our complex but static world. On the positive side, we find that the learning process allows cooperative systems to coordinate around satisficing solutions with rather high (but markedly sub-optimal) average reward. However, hyper-sensitivity to the game parameters makes it impossible to predict ex ante who will be better or worse off in our stylized economy.

arXiv comment35 pages, 19 figures

References in corpus (53)

• A survey of random processes with reinforcement
• On the Out of Equilibrium Relaxation of the Sherrington - Kirkpatrick model
• Stimulus-Dependent Suppression of Chaos in Recurrent Neural Networks
• Nonequilibrium glass transitions in driven and active matter
• Crises and collective socio-economic phenomena: simple models and challenges
• Marginal Stability in Structural, Spin and Electron Glasses
• Geometrical Aspects of Aging and Rejuvenation in the Ising Spin Glass: A Numerical Study
• Transition to chaos in random neuronal networks
• Numerical implementation of dynamical mean field theory for disordered systems: application to the Lotka-Volterra model of ecosystems
• Nonequilibrium Dynamics and Aging in the Three--Dimensional Ising Spin Glass Model
• Nonequilibrium Dynamics and Aging in the Three--Dimensional Ising Spin Glass Model
• Rational Decisions, Random Matrices and Spin Glasses
• Dynamics of Random Neural Networks with Bistable Units
• Properties of equilibria and glassy phases of the random Lotka-Volterra model with demographic noise
• Glassy behaviour in disordered systems with non-relaxational dynamics
• Dynamics of Boltzmann Q-Learning in Two-Player Two-Action Games
• Explosive cooperation in social dilemmas on higher-order networks
• Correlations between synapses in pairs of neurons slow down dynamics in randomly connected neural networks
• Stability and Diversity in Collective Adaptation
• Path Integral Approach to Random Neural Networks
• Evidence of aging in mean-field spin glass models
• The Complexity of Ising Spin Glasses
• May's Instability in Large Economies
• The prevalence of chaotic dynamics in games with many players
• Intrinsic noise in game dynamical learning
• Generalized Lotka-Volterra equations with random, non-reciprocal interactions: the typical number of equilibria
• Random replicators with asymmetric couplings
• Low temperature solution of the Sherrington-Kirkpatrick model
• Disorder chaos in spin glasses
• Chaos and unpredictability in evolutionary dynamics in discrete time
• When can the two-armed bandit algorithm be trusted?
• Off-Equilibrium Dynamics of 1+1 dimensional Directed Polymer in Random Media
• Ground state of the Bethe-lattice spin glass and running time of an exact optimization algorithm
• Free energy landscapes, dynamics and the edge of chaos in mean-field models of spin glasses
• The phase diagram of Lévy spin glasses
• Cycles of cooperation and defection in imperfect learning
• Simulations of Ground State Fluctuations in Mean-Field Ising Spin Glasses
• By Force of Habit: Self-Trapping in a Dynamical Utility Landscape
• Relaxation, closing probabilities and transition from oscillatory to chaotic attractors in asymmetric neural networks
• Numerical Simulations of the Dynamical Behavior of the SK Model
• On the number of limit cycles in asymmetric neural networks
• Thermodynamics of the Lévy spin glass
• Aging by near-extinctions in many-variable interacting populations
• Bond chaos in the Sherrington-Kirkpatrick model
• On the Stability of the Mean-Field Glass Broken Phase under Non-Hamiltonian Perturbations
• Collective states in social systems with interacting learning agents
• Good speciation and endogenous business cycles in a constraint satisfaction macroeconomic model
• The realizable solutions of the TAP equations
• A new spin on optimal portfolios and ecological equilibria
• Self-Excited Ising Game
• Off-Equilibrium Dynamics of a 4D Spin Glass with Asymmetric Couplings
• Memory and limit cycles in rock-scissors-paper
• A model with Darwinian dynamics on a rugged landscape

Cited by in corpus (8)

• Nonreciprocal Ising model
• Dynamical phase transitions in the nonreciprocal Ising model
• Nonreciprocal Spin-Glass Transition and Aging
• Quenches in the Sherrington-Kirkpatrick model
• Dynamical Theory for Adaptive Systems
• Irreversibility in Non-reciprocal Chaotic Systems
• Phase Transitions in single species Ising Models with Non-Reciprocal couplings
• Zero-temperature dynamics of the spherical model with non-reciprocal interactions