In my research, I use an interdisciplinary approach that combines Behavioral and Experimental Economics, Cognitive Psychology, Cognitive Neuroscience, and Psychophysics. I am particularly interested in how attentional mechanisms, logical thinking, and reasoning skills are involved in economic decision-making. Supervisors: Aldo Rustichini, Giorgio Coricelli, and Nicolao Bonini Phone: Mobile: +39-3881990266 Address: N. 9, via C. Battisti, 35047, Solesino, Padova, Italy
Proceedings of the ACM on Human-Computer Interaction
Image classification models are becoming a popular method of analysis for scanpath classification... more Image classification models are becoming a popular method of analysis for scanpath classification. To implement these models, gaze data must first be reconfigured into a 2D image. However, this step gets relatively little attention in the literature as focus is mostly placed on model configuration. As standard model architectures have become more accessible to the wider eye-tracking community, we highlight the importance of carefully choosing feature representations within scanpath images as they may heavily affect classification accuracy. To illustrate this point, we create thirteen sets of scanpath designs incorporating different eye-tracking feature representations from data recorded during a task-based viewing experiment. We evaluate each scanpath design by passing the sets of images through a standard pre-trained deep learning model as well as a SVM image classifier. Results from our primary experiment show an average accuracy improvement of 25 percentage points between the bes...
Eye movement data has been extensively utilized by researchers interested in studying decision-ma... more Eye movement data has been extensively utilized by researchers interested in studying decision-making within the strategic setting of economic games. In this paper, we demonstrate both a deep learning and traditional machine learning classification method which are able to accurately identify a given participant's decision strategy before they commit to an action while playing games. Our approach focuses on creating scanpath images that best capture the dynamics of a participant's gaze behaviour during a given game in a way that is meaningful to the machine learning models. Our results demonstrate a higher classification accuracy compared to traditional methods of analysis applied to the same economic game environments by as much as 18 percentage points. In a broader context, we aim to illustrate the potential for eye-tracking data to create information asymmetries in strategic environments in favour of those who collect and process the data. These information asymmetries co...
Abstract We propose an experimental eye-tracking study to test how strategic sophistication is sh... more Abstract We propose an experimental eye-tracking study to test how strategic sophistication is shaped by experience in 3 × 3 two-person normal-form games. Although strategic sophistication has been shown to be linked to a variety of endogenous and exogenous factors, little is known about how it is affected by previous interactive decisions. We show that complete feedback in previous games can significantly enhance strategic sophistication, and that games that in principle provide equivalent learning opportunities lead instead to substantially different learning outcomes. Specifically, only repeated play with feedback of games that emphasize strategic interdependence significantly enhances strategic learning, producing an increase in the frequency of equilibrium play and a shift of attention to the incentives of the counterpart. Moreover, we find that the type of learning underlying newly gained strategic skills can vary substantially across players. Whereas some players eventually learn to visually analyze the payoff matrix consistently with equilibrium reasoning, others appear to use experience with previous interactions to devise simple heuristics of play. Our results have implications for theoretical and computational modeling of learning.
In a series of three behavioral experiments, we found a systematic distortion of probability judg... more In a series of three behavioral experiments, we found a systematic distortion of probability judgments concerning elementary visual stimuli. Participants were briefly shown sets of figures that had two features (e.g., a geometric shape and a color) with two possible values (e.g., triangle or circle and black or white). For each set, a figure was drawn, participants were informed about the value of one of its features (e.g., that the figure was a “circle”) and had to predict the value of the other (e.g., whether the figure was “black” or “white”). By varying the statistical association between features in the various sets of figures, we manipulated the probability of a feature given the evidence of another (e.g., the posterior probability of hypothesis “black” given the evidence “circle”) as well as the support provided by a feature to another (e.g., the impact, or confirmation, of evidence “circle” on the hypothesis “black”). Results indicated that participants’ judgments were deepl...
In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior... more In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior based on the possible actions of others. Growing experimental evidence has shown that players implement different levels of sophistication in games, as described by hierarchical models of strategic thinking such as Level-k and Cognitive Hierarchy. In the current study, we explore the role of two cognitive factors – game representation and cognitive reflection - in explaining heterogeneity in strategic behavior. In particular, we hypothesize that higher levels of cognitive reflection specifically predict the implementation of more sophisticated processes of game representation, which in turn explains higher levels of strategic sophistication. In two eye-tracking experiments, we registered eye movements of participants while playing matrix games of increasing relational complexity (2x2 and 3x3 matrices), and we analyzed individual patterns of information acquisition to reveal the ongoing process of game representation-building. In order to assess individual levels of cognitive reflection, participants completed the Cognitive Reflection Test (CRT), among different control measures of cognitive ability. Results show that, in both classes of games, higher cognitive reflection levels specifically predicted the ability to incorporate the counterpart’s incentives in the model of the current game, as well as higher levels of strategic sophistication. Conversely, players exhibiting low cognitive reflection disregarded relevant transitions between the payoffs of the counterpart, and such incomplete visual analysis led to out-of-equilibrium choices. The relationship between cognitive reflection and strategic choices was completely mediated by gaze patterns. Our results shed new light on the cognitive factors driving heterogeneity in strategic thinking, and on theories of bounded rationality.
Growing evidence in behavioral decision making supports the existence of two distinct decision st... more Growing evidence in behavioral decision making supports the existence of two distinct decision strategies in uncertain environments, a model-based and a model-free approach. The former relies on the construction of an exhaustive model of the environment, while the latter does not grasp the relational complexity underlying contingencies. However, no studies have investigated whether model-based and model-free behaviors express distinct processes of representations generation. In the present study, we investigated the process of representation-building by tracking eye movements of participants who performed a novel relational reasoning task. A cluster analysis on gaze data confirmed the existence of two distinct types of participant that respectively expressed model-based and model-free representation processes. Model-based agents systematically searched for the higher-order relations characterizing the environment, while model-free agents encoded simple rules without exploring the underlying relational complexity. Analyses of individual cognitive measures revealed that cognitive reflection is associated with the emergence of either model-based or model-free behavior, while working memory and fluid intelligence sustain maintenance and updating functio ns. Our results reveal the existence of two distinct processes of representation generation and provide new insights about the cognitive mechanisms underlying model-based and model-free behaviors.
Proceedings of the ACM on Human-Computer Interaction
Image classification models are becoming a popular method of analysis for scanpath classification... more Image classification models are becoming a popular method of analysis for scanpath classification. To implement these models, gaze data must first be reconfigured into a 2D image. However, this step gets relatively little attention in the literature as focus is mostly placed on model configuration. As standard model architectures have become more accessible to the wider eye-tracking community, we highlight the importance of carefully choosing feature representations within scanpath images as they may heavily affect classification accuracy. To illustrate this point, we create thirteen sets of scanpath designs incorporating different eye-tracking feature representations from data recorded during a task-based viewing experiment. We evaluate each scanpath design by passing the sets of images through a standard pre-trained deep learning model as well as a SVM image classifier. Results from our primary experiment show an average accuracy improvement of 25 percentage points between the bes...
Eye movement data has been extensively utilized by researchers interested in studying decision-ma... more Eye movement data has been extensively utilized by researchers interested in studying decision-making within the strategic setting of economic games. In this paper, we demonstrate both a deep learning and traditional machine learning classification method which are able to accurately identify a given participant's decision strategy before they commit to an action while playing games. Our approach focuses on creating scanpath images that best capture the dynamics of a participant's gaze behaviour during a given game in a way that is meaningful to the machine learning models. Our results demonstrate a higher classification accuracy compared to traditional methods of analysis applied to the same economic game environments by as much as 18 percentage points. In a broader context, we aim to illustrate the potential for eye-tracking data to create information asymmetries in strategic environments in favour of those who collect and process the data. These information asymmetries co...
Abstract We propose an experimental eye-tracking study to test how strategic sophistication is sh... more Abstract We propose an experimental eye-tracking study to test how strategic sophistication is shaped by experience in 3 × 3 two-person normal-form games. Although strategic sophistication has been shown to be linked to a variety of endogenous and exogenous factors, little is known about how it is affected by previous interactive decisions. We show that complete feedback in previous games can significantly enhance strategic sophistication, and that games that in principle provide equivalent learning opportunities lead instead to substantially different learning outcomes. Specifically, only repeated play with feedback of games that emphasize strategic interdependence significantly enhances strategic learning, producing an increase in the frequency of equilibrium play and a shift of attention to the incentives of the counterpart. Moreover, we find that the type of learning underlying newly gained strategic skills can vary substantially across players. Whereas some players eventually learn to visually analyze the payoff matrix consistently with equilibrium reasoning, others appear to use experience with previous interactions to devise simple heuristics of play. Our results have implications for theoretical and computational modeling of learning.
In a series of three behavioral experiments, we found a systematic distortion of probability judg... more In a series of three behavioral experiments, we found a systematic distortion of probability judgments concerning elementary visual stimuli. Participants were briefly shown sets of figures that had two features (e.g., a geometric shape and a color) with two possible values (e.g., triangle or circle and black or white). For each set, a figure was drawn, participants were informed about the value of one of its features (e.g., that the figure was a “circle”) and had to predict the value of the other (e.g., whether the figure was “black” or “white”). By varying the statistical association between features in the various sets of figures, we manipulated the probability of a feature given the evidence of another (e.g., the posterior probability of hypothesis “black” given the evidence “circle”) as well as the support provided by a feature to another (e.g., the impact, or confirmation, of evidence “circle” on the hypothesis “black”). Results indicated that participants’ judgments were deepl...
In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior... more In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior based on the possible actions of others. Growing experimental evidence has shown that players implement different levels of sophistication in games, as described by hierarchical models of strategic thinking such as Level-k and Cognitive Hierarchy. In the current study, we explore the role of two cognitive factors – game representation and cognitive reflection - in explaining heterogeneity in strategic behavior. In particular, we hypothesize that higher levels of cognitive reflection specifically predict the implementation of more sophisticated processes of game representation, which in turn explains higher levels of strategic sophistication. In two eye-tracking experiments, we registered eye movements of participants while playing matrix games of increasing relational complexity (2x2 and 3x3 matrices), and we analyzed individual patterns of information acquisition to reveal the ongoing process of game representation-building. In order to assess individual levels of cognitive reflection, participants completed the Cognitive Reflection Test (CRT), among different control measures of cognitive ability. Results show that, in both classes of games, higher cognitive reflection levels specifically predicted the ability to incorporate the counterpart’s incentives in the model of the current game, as well as higher levels of strategic sophistication. Conversely, players exhibiting low cognitive reflection disregarded relevant transitions between the payoffs of the counterpart, and such incomplete visual analysis led to out-of-equilibrium choices. The relationship between cognitive reflection and strategic choices was completely mediated by gaze patterns. Our results shed new light on the cognitive factors driving heterogeneity in strategic thinking, and on theories of bounded rationality.
Growing evidence in behavioral decision making supports the existence of two distinct decision st... more Growing evidence in behavioral decision making supports the existence of two distinct decision strategies in uncertain environments, a model-based and a model-free approach. The former relies on the construction of an exhaustive model of the environment, while the latter does not grasp the relational complexity underlying contingencies. However, no studies have investigated whether model-based and model-free behaviors express distinct processes of representations generation. In the present study, we investigated the process of representation-building by tracking eye movements of participants who performed a novel relational reasoning task. A cluster analysis on gaze data confirmed the existence of two distinct types of participant that respectively expressed model-based and model-free representation processes. Model-based agents systematically searched for the higher-order relations characterizing the environment, while model-free agents encoded simple rules without exploring the underlying relational complexity. Analyses of individual cognitive measures revealed that cognitive reflection is associated with the emergence of either model-based or model-free behavior, while working memory and fluid intelligence sustain maintenance and updating functio ns. Our results reveal the existence of two distinct processes of representation generation and provide new insights about the cognitive mechanisms underlying model-based and model-free behaviors.
We propose an experimental eye-tracking study to test how strategic sophistication is shaped by e... more We propose an experimental eye-tracking study to test how strategic sophistication is shaped by experience in 3x3 two-person normal-form games. Although strategic sophistication has been shown to be linked to a variety of endogenous and exogenous factors, little is known about how it is affected by previous interactive decisions. We show that complete feedback in previous games can significantly enhance strategic sophistication, and that games that in principle provide equivalent learning opportunities lead instead to substantially different learning outcomes. Specifically, only repeated play with feedback of games that emphasize strategic interdependence significantly enhances strategic learning, producing an increase in the frequency of equilibrium play and a shift of attention to the incentives of the counterpart. Moreover, we find that the type of learning underlying newly gained strategic skills can vary substantially across players. Whereas some players eventually learn to visually analyze the payoff matrix consistently with equilibrium reasoning, others appear to use experience with previous interactions to devise simple heuristics of play. Our results have implications for theoretical and computational modeling of learning.
In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior... more In social contexts, we refer to strategic sophistication as the ability to adapt our own behavior based on the possible actions of others. Growing experimental evidence has shown that players implement different levels of sophistication in games, as described by hierarchical models of strategic thinking such as Level-k and Cognitive Hierarchy. In the current study, we explore the role of two cognitive factors – game representation and cognitive reflection - in explaining heterogeneity in strategic behavior. In particular, we hypothesize that higher levels of cognitive reflection specifically predict the implementation of more sophisticated processes of game representation, which in turn explains higher levels of strategic sophistication. In two eye-tracking experiments, we registered eye movements of participants while playing matrix games of increasing relational complexity (2x2 and 3x3 matrices), and we analyzed individual patterns of information acquisition to reveal the ongoing process of game representation-building. In order to assess individual levels of cognitive reflection, participants completed the Cognitive Reflection Test (CRT), among different control measures of cognitive ability. Results show that, in both classes of games, higher cognitive reflection levels specifically predicted the ability to incorporate the counterpart’s incentives in the model of the current game, as well as higher levels of strategic sophistication. Conversely, players exhibiting low cognitive reflection disregarded relevant transitions between the payoffs of the counterpart, and such incomplete visual analysis led to out-of-equilibrium choices. The relationship between cognitive reflection and strategic choices was completely mediated by gaze patterns. Our results shed new light on the cognitive factors driving heterogeneity in strategic thinking, and on theories of bounded rationality.
In our everyday life, we often need to anticipate the potential occurrence of events and their re... more In our everyday life, we often need to anticipate the potential occurrence of events and their respective consequences. In this context, the way we represent contingencies can determine our ability to adapt to the environment. However, it is not clear how agents encode and organize relevant information to react to possible states of the world. In the present study, we investigated the process of representation-building by tracking eye movements of participants who performed a novel relational inference task, where conditional rules regulated the occurrence of interdependent events. A cluster analysis on early gaze data revealed the existence of two distinct types of participants. A group of (sophisticated) participants systematically searched for the higher-order relations characterizing the environment, building an exhaustive model of the underlying contingencies. Another group of (unsophisticated) participants simply learned binary conditional rules without exploring the underlying relational complexity, trying to infer the indirect consequences of an event only after its occurrence. Analyses of individual cognitive measures revealed that cognitive reflection is associated with the emergence of either sophisticated or unsophisticated representation behavior, while working memory and fluid intelligence sustain maintenance and updating functions. Our results reveal the existence of two distinct spontaneous processes of representation generation, which might explain part of the heterogeneity observed in several experimental fields, including learning, decision-making and reasoning.
This chapter is about the application of methods from psychophysics to the investigation of exper... more This chapter is about the application of methods from psychophysics to the investigation of experimental game theory. We show how methods from psychophysics-including the analysis of reaction time, mouse-tracking and eye-tracking,-can be used to improve our understanding of experimental game theory. The main goal of this chapter is to provide a balanced view of the possibilities of the process tracing approach. In addition, to provide enough practical understanding of the methods to enable the reader evaluate the literature that is increasingly growing. A secondary goal is to provide an introduction for readers interested in designing their own psychophysics experiments.
We study experimentally how people learn from observing the choices of others in a non-stationary... more We study experimentally how people learn from observing the choices of others in a non-stationary stochastic environment. The imitation choices of participants with low score in an intelligence test are driven solely by the value of imitation. High intelligence score participants , in addition, use choices of others to better understand the environment. They imitate more when other's choices are stable, which makes them more optimal than low score participants. The knowledge that the other has high intelligence score increases the optimality of only low score participants. Overall, intelligence predicts the usage of simple or sophisticated observational learning strategy. JEL classifications: C91, C92
We use eye-tracking technique to test whether players’ actions are consistent with their expectat... more We use eye-tracking technique to test whether players’ actions are consistent with their expectations of their opponent’s behavior. Participants play a series of two-player 3 by 3 one shot games and state their beliefs about which actions they expect their counterpart to play (first-order beliefs) or about which actions their counterparts expect them to play (second-order beliefs). We perform a mixed model cluster analysis and classify participants into types according to both their information search patterns and choices. Players classified as strategic (Level-2 players) and players classified as having other regarding preferences like Inequity aversion and Prosociality exhibit attentional patterns of visual information acquisition and choices that are mainly consistent with their stated beliefs. Conversely, players classified as non-strategic (Level-1, Pessimistic, Optimistic and Competitive) do not best respond to any specific belief, but apply simple decision rules regardless of whether they are playing or stating their beliefs. Thus, using eye-tracking data we could identify a larger consistency between actions and stated beliefs compared with previous studies, and we could characterize the behavioral rules associated with choice-beliefs inconsistency. Implications for the theories of bounded rationality are discussed.
We used eye-tracking to measure the dynamic patterns of visual information acquisition in two pla... more We used eye-tracking to measure the dynamic patterns of visual information acquisition in two players normal form games. Participants played one-shot games in which either, neither, or only one of the players had a dominant strategy. First, we performed a mixture models cluster analysis to group participants into types according to the pattern of visual information acquisition observed in a single class of games. Then, we predicted agents’ choices in different classes of games, and observed that patterns of visual information acquisition were game invariant. Our method allowed us to predict whether the decision process would lead to equilibrium choices or not, and to attribute out-of-equilibrium responses to limited cognitive capacities or social motives. Our results suggest the existence of individually heterogeneous-but stable-patterns of visual information acquisition based on subjective levels of strategic sophistication and social preferences.
We used eye-tracking to measure the dynamic patterns of visual information acquisition in two pla... more We used eye-tracking to measure the dynamic patterns of visual information acquisition in two players normal form games. Participants played one-shot games in which either, neither, or only one of the players had a dominant strategy. First, we performed a mixture models cluster analysis to group participants into types according to the pattern of visual information acquisition observed in a single class of games. Then, we predicted agents’ choices in different classes of games, and observed that patterns of visual information acquisition were game invariant. Our method allowed us to predict whether the decision process would lead to equilibrium choices or not, and to attribute out-ofequilibrium responses to limited cognitive capacities or social motives. Our results suggest the existence of individually heterogeneous-but stable-patterns of visual information acquisition based on subjective levels of strategic sophistication and social preferences.
"Objective: Most models of choices assume that a player’s actions are guided by beliefs. Neverthe... more "Objective: Most models of choices assume that a player’s actions are guided by beliefs. Nevertheless, many studies found choice data are largely inconsistent with belief data. We test the hypothesis of consistency between beliefs and choices, and that the visual patterns of information's acquisition can be exploited to predict both players’ choices and beliefs.
Methods: We recorded eye movements of 100 participants playing a set of 18 two person 3 by 3 games and their stated beliefs about which actions they expected their counterpart to choose. The games included dominant solvable games, pure Nash equilibrium games and coordination games. The games were chosen to achieve a strong separation between 9 models of game play: Equilibrium, Level-2, Dominance-1, Level-1/Naïve, Optimistic, Pessimistic, Inequity Averse, Prosocial and Competitive. We then performed a mixture model cluster analysis to group participants into types according to the prevailing payoff comparisons they made when visually analyzing the games in the choice task.
Results: The cluster analysis differentiated between 3 general types of visual analysis, according to which we categorized players into different types: one cluster of players was shown to focus prevalently on information regarding their own payoffs, and they chose accordingly with 3 simple non-strategic models (Level-1, Pessimistic and Optimistic). A second cluster contained players focused mainly on the comparison between their own and their counterpart payoffs. They chose accordingly with 3 models of social value orientation (Inequity Averse, Prosocial and Competitive). The third cluster contained players which exhibited a more balanced mixture of all types of visual analysis patterns, and they chose accordingly with strategic models (Equilibrium and level-2).
Players classified as either Equilibrium, Level-2, Inequity Averse, Prosocial or Competitive exhibited patterns of visual analysis and choices that were mainly consistent with their stated beliefs regarding the choice of their counterpart. Conversely, the beliefs of players classified as Level-1, Pessimistic or Optimistic were inconsistent with their visual analysis and choices. This suggests that the latter players did not form beliefs at all and used simplifying heuristics.
Conclusion: We show that it is possible to predict players’ choices and beliefs by knowing how they visually analyze interactive decision problems in game playing. Finally, considering the existence of individually heterogeneous patterns of responses, which are based on a set of a priori types, we found the players’ behavior to be generally consistent with their beliefs."
Game theory proposes that optimal interactive decision making requires subjects to employ differe... more Game theory proposes that optimal interactive decision making requires subjects to employ different strategies according to the game. In particular, deliberative processes are necessary to solve dominant solvable games, but are inefficient in coordination games, where intuition is required. We aimed to link distinct visual analysis patterns of games to these two processes. To do so we used Eye-tracking to examine how subjects visually analyzed games and investigated whether this would predict subsequent choices. We clustered subjects into “types” according to the prevailing payoff comparisons they made in their visual analyses. This resulted in 3 general types of players: Altruistic players, which focus their attention mostly on possible game outcomes; Own focused players, which focus their attention prevalently on their own payoffs; and strategic players, which employed a relatively balanced mixture of visual analyses types. Although clustering was performed on 1 type of game, it predicted players’ responses in all of the other ones. We suggest this might be due to limited visual analysis patterns, which could lead to misrepresenting the games. Altruistic players’ visual patterns suggest they are using intuition. This may drive the appropriate strategy selection in stag hunt games but may induce them towards out of equilibrium strategies in dominant solvable games. Conversely, strategic players’ visual analyses suggest they are using deliberative processes in solving the games. This is supported by the observation that they detect dominant strategies when present, but appear unable to identify the possibility to coordinate in stag hunt games. Own focused players almost never try to predict their counterparts’ responses and appear to follow elementary heuristics.
"Luca Polonio; Giorgio Coricelli; Nicolao Bonini.
Objective: Standard game theory makes no exp... more "Luca Polonio; Giorgio Coricelli; Nicolao Bonini.
Objective: Standard game theory makes no explicit claims about mental processes underlying choice. Our objective is to directly investigate if it is possible to predict players’ behavior in different types of games by analyzing their own and their counterparts’ payoffs in a single type of game.
Methods: We recorded eye-movements of 60 participants playing 32 randomly presented games.
Games were divided into 4 classes requiring distinct strategies: dominance solvable self games, in which the participant had himself a dominant strategy; dominant solvable other games, in which the participants’ counterpart had a dominant strategy; prisoner’s dilemma games, which elicit conflict between a dominant strategy and social cooperation; and stag-hunt games, which elicit conflict between a safety choice and a risky but possibly higher paying one. We then performed mixture models cluster analysis to group participants into types according to the prevailing payoff comparisons they made when visually analyzing the game. Only one class of games was used in this analysis. We finally used the same clusters to predict participants’ choices in the other games.
Results: The cluster analysis differentiated between 3 general types of analysis, according to which we categorized players into distinct types: one type of players (own focused players) was shown to focus prevalently on information regarding their own payoffs. A second type (altruistic players) focused mainly on possible game results. The third type (strategic players) contained players which exhibited a more balanced mixture of all types of visual analysis patterns. Such visual analysis based categorizations predicted the participants’ choices throughout all games. Altruistic players easily detected Pareto equilibrium in stag-hunt games but failed to choose equilibrium strategies in dominant solvable games. Conversely, strategic players were able to detect equilibrium in dominant solvable games but almost never chose the Pareto superior solution in the stag-hunt games. Own focused players responded according with the equilibrium when they had a dominant strategy, but not when only their counterparts did. They moreover chose the safe solution in the stag-hunt games.
Conclusions: We show that it is possible to predict subjects’ choices by examining the way they visually analyze interactive decision problems. Our experiment found players’ behavior to be generally suboptimal, as they appeared unable to change their strategies according to the game. On the basis of our results we suggest that this might be due to limited visual analysis patterns, which, in turn, could lead to misrepresentations of the games."
This chapter is about the application of methods from psychophysics to the investigation of exper... more This chapter is about the application of methods from psychophysics to the investigation of experimental game theory. We show how methods from psychophysics-including the analysis of reaction time, mouse-lab and eye-tracking,-can be used to improve our understanding of experimental game theory. The main goal of this chapter is to provide a balanced view of the possibilities of the process tracing approach, which is the investigation of the processes underlying choice. In addition, this chapter aims to provide practical understanding of the methods to enable the reader to evaluate this fast-growing literature. A secondary goal is to provide an introduction for readers interested in designing their own psychophysics experiments.
Spesso, investitori ed economisti si rivolgono alla teoria dei giochi per aiutarsi a prendere sce... more Spesso, investitori ed economisti si rivolgono alla teoria dei giochi per aiutarsi a prendere scelte ottimali, ma questa può spesso indurre ad errori. Infatti, la teoria dei giochi fa previsioni assumendo capacità cognitive illimitate, che sono implausibili dal punto di vista cognitivo, e che potrebbero specialmente essere fuorvianti laddove le nostre scelte dipendono da altri, che potrebbero non comportarsi in mo-do poi così strategico. Durante gli ultimi vent'anni, le neuroscienze e le scienze cognitive hanno dipinto un'immagine più psico-biologicamente plausibile di come ragionano e interagiscono gli esseri umani. Nella prima parte di questo capitolo, de-scriveremo come una rete cerebrale particolare, incentrata sulla corteccia mediale prefrontale, sembri giocare un ruolo particolare nella cognizione sociale (e non necessariamente nella cognizione in generale) e nella sofisticatezza del ragionamento strategico in un concorso finanziario ispirato da Keynes. Quest'area cerebrale potrebbe essere specializzata nel compiere inferenze sugli altri, mentre questi compiono inferenze su di noi. Nella seconda parte, illustreremo diversi modi in cui, la mente-cervello potrebbe aver tentato di semplificare i problemi decisionali, portandoci a restringere l'attenzione su alcune informazioni, a scapito di altre. Infatti, mostreremo come molte delle nostre decisioni potrebbero già essere racchiuse nel modo in cui analizziamo visivamente le informazioni a disposizione. Plausibilmente, una futura teoria delle decisioni riuscirà ad integrare la teoria dei giochi, con una teoria della mente. " Gli investimenti finanziari dei professionisti assomigliano a quelle competizioni che si trovano nei giornali, nelle quali i concorrenti devono scegliere i volti più belli tra un centinaio di fotografie e il premio viene assegnato al concorrente la
Uploads
Papers by luca polonio
Results show that, in both classes of games, higher cognitive reflection levels specifically predicted the ability to incorporate the counterpart’s incentives in the model of the current game, as well as higher levels of strategic sophistication. Conversely, players exhibiting low cognitive reflection disregarded relevant transitions between the payoffs of the counterpart, and such incomplete visual analysis led to out-of-equilibrium choices. The relationship between cognitive reflection and strategic choices was completely mediated by gaze patterns. Our results shed new light on the cognitive factors driving heterogeneity in strategic thinking, and on theories of bounded rationality.
Methods: We recorded eye movements of 100 participants playing a set of 18 two person 3 by 3 games and their stated beliefs about which actions they expected their counterpart to choose. The games included dominant solvable games, pure Nash equilibrium games and coordination games. The games were chosen to achieve a strong separation between 9 models of game play: Equilibrium, Level-2, Dominance-1, Level-1/Naïve, Optimistic, Pessimistic, Inequity Averse, Prosocial and Competitive. We then performed a mixture model cluster analysis to group participants into types according to the prevailing payoff comparisons they made when visually analyzing the games in the choice task.
Results: The cluster analysis differentiated between 3 general types of visual analysis, according to which we categorized players into different types: one cluster of players was shown to focus prevalently on information regarding their own payoffs, and they chose accordingly with 3 simple non-strategic models (Level-1, Pessimistic and Optimistic). A second cluster contained players focused mainly on the comparison between their own and their counterpart payoffs. They chose accordingly with 3 models of social value orientation (Inequity Averse, Prosocial and Competitive). The third cluster contained players which exhibited a more balanced mixture of all types of visual analysis patterns, and they chose accordingly with strategic models (Equilibrium and level-2).
Players classified as either Equilibrium, Level-2, Inequity Averse, Prosocial or Competitive exhibited patterns of visual analysis and choices that were mainly consistent with their stated beliefs regarding the choice of their counterpart. Conversely, the beliefs of players classified as Level-1, Pessimistic or Optimistic were inconsistent with their visual analysis and choices. This suggests that the latter players did not form beliefs at all and used simplifying heuristics.
Conclusion: We show that it is possible to predict players’ choices and beliefs by knowing how they visually analyze interactive decision problems in game playing. Finally, considering the existence of individually heterogeneous patterns of responses, which are based on a set of a priori types, we found the players’ behavior to be generally consistent with their beliefs."
Objective: Standard game theory makes no explicit claims about mental processes underlying choice. Our objective is to directly investigate if it is possible to predict players’ behavior in different types of games by analyzing their own and their counterparts’ payoffs in a single type of game.
Methods: We recorded eye-movements of 60 participants playing 32 randomly presented games.
Games were divided into 4 classes requiring distinct strategies: dominance solvable self games, in which the participant had himself a dominant strategy; dominant solvable other games, in which the participants’ counterpart had a dominant strategy; prisoner’s dilemma games, which elicit conflict between a dominant strategy and social cooperation; and stag-hunt games, which elicit conflict between a safety choice and a risky but possibly higher paying one. We then performed mixture models cluster analysis to group participants into types according to the prevailing payoff comparisons they made when visually analyzing the game. Only one class of games was used in this analysis. We finally used the same clusters to predict participants’ choices in the other games.
Results: The cluster analysis differentiated between 3 general types of analysis, according to which we categorized players into distinct types: one type of players (own focused players) was shown to focus prevalently on information regarding their own payoffs. A second type (altruistic players) focused mainly on possible game results. The third type (strategic players) contained players which exhibited a more balanced mixture of all types of visual analysis patterns. Such visual analysis based categorizations predicted the participants’ choices throughout all games. Altruistic players easily detected Pareto equilibrium in stag-hunt games but failed to choose equilibrium strategies in dominant solvable games. Conversely, strategic players were able to detect equilibrium in dominant solvable games but almost never chose the Pareto superior solution in the stag-hunt games. Own focused players responded according with the equilibrium when they had a dominant strategy, but not when only their counterparts did. They moreover chose the safe solution in the stag-hunt games.
Conclusions: We show that it is possible to predict subjects’ choices by examining the way they visually analyze interactive decision problems. Our experiment found players’ behavior to be generally suboptimal, as they appeared unable to change their strategies according to the game. On the basis of our results we suggest that this might be due to limited visual analysis patterns, which, in turn, could lead to misrepresentations of the games."