Publications

Game-theoretic models of learning are hard to study even in the laboratory setting due to econometric and practical concerns (like the limited length of an experimental session). In particular, as the simulations by (Salmon, 2001) show, in a cross-model (or "blind'') testing of several models, the data generated by those models does not correspond to the estimated parameters correctly. Thus, even when the real data generation process is known we cannot distinguish correct models from incorrect ones by looking at the estimates. However, we demonstrate that under the same conditions, models are clearly distinguishable if we compare predictions that the models make instead of comparing the model parameters. We also provide a rationale for why this cross-model predictive quality is a particularly relevant way for improving learning models.

We systematically investigate prisoner’s dilemma and dictator games with valence framing. We find that give versus take frames influence subjects’ behavior and beliefs in the prisoner’s dilemma games but not in the dictator games. We conclude that valence framing has a stronger impact on behavior in strategic interactions, i.e., in the prisoner’s dilemma game, than in allocation tasks without strategic interaction, i.e., in the dictator game.

This paper studies learning in strategic environment using experimental data from the Rock-Paper-Scissors game. In a repeated game framework, we explore the response of human subjects to uncertain behavior of strategically sophisticated opponent. We model this opponent as a robot who played a stationary strategy with superimposed noise varying across four experimental treatments. Using experimental data from 85 subjects playing against such a stationary robot for 100 periods, we show that humans can decode their strategies, on average outperforming the random response to such a robot by 17%. Further, we show that human ability to recognize such strategies decreases with exogenous noise in the behavior of the robot. Further, we fit learning data to classical Reinforcement Learning (RL) and Fictitious Play (FP) models and show that the classic action-based approach to learning is inferior to the strategy-based one. Unlike the previous papers in this field, e.g. Ioannou, Romero (2014), we extend and adapt the strategy-based learning techniques to the 3x3 game. We also show, using a combination of experimental and ex-post survey data, that human participants are better at learning separate components of an opponent's strategy than in recognizing this strategy as a whole. This decomposition offers them a shorter and more intuitive way to figure out their own best response. We build a strategic extension of the classical learning models accounting for these behavioral phenomena.

The history of research in finance and economics has been widely impacted by the field of Agent-based Computational Economics (ACE). While at the same time being popular among natural science researchers for its proximity to the successful methods of physics and chemistry for example, the field of ACE has also received critics by a part of the social science community for its lack of empiricism. Yet recent trends have shifted the weights of these general arguments and poten- tially given ACE a whole new range of realism. At the base of these trends are found two present-day major scientific breakthroughs: the steady shift of psychology towards a hard science due to the advances of neuropsychology, and the progress of reinforcement learning due to increasing computational power and big data. We outline here the main lines of a computational research study where each agent would trade by reinforcement learning.

According to theories on moral balancing, a prosocial act can decrease people’s motivation to engage in subsequent prosocial behavior, because people feel that they have already achieved a positive moral self-perception. However, there is also empirical evidence showing that people actually need to be recognized by others in order to establish and affirm their self-perception through their prosocial actions. Without social recognition, moral balancing could possibly fail. In this paper, we investigate in two laboratory experiments how social recognition of prosocial behavior influences subsequent moral striving. Building on self-completion theory, we hypothesize that social recognition of prosocial behavior (self-serving behavior) weakens (strengthens) subsequent moral striving. In Study 1, we show that a prosocial act leads to less subsequent helpfulness when it was socially recognized as compared to a situation without social recognition. Conversely, when a self-serving act is socially recognized, it encourages subsequent helpfulness. In Study 2, we replicate the effect of social recognition on moral striving in a more elaborated experimental setting and with a larger participant sample. We again find that a socially recognized prosocial act leads to less subsequent helpfulness compared to an unrecognized prosocial act. Our results shed new light on the boundary conditions of moral balancing effects and underscore the view that these effects can be conceptualized as a dynamic of self-completion.

We analyze the effects of limited feedback on beliefs and contributions in a repeated public goods game setting. In a first experiment, we test whether exogenously determined feedback about a good example (i.e., the maximum con- tribution in a period) in contrast to a bad example (i.e., the minimum contribution in a period) induces higher contributions. We find that when the type of feedback is not transparent to the group members, good examples boost cooperation while bad examples hamper it. There is no difference when the type of feedback is transparent. In a second experiment, feedback is endogenously chosen by a group leader. The results show that a large majority of the group leaders count on the positive effect of providing a good example. This is true regardless whether they choose the feedback type to be transparent or non-transparent. Half of the group leaders make the type of feedback transparent. With endogenously chosen feedback about good examples no difference in contributions can be observed among transparent and non-transparent feedback selection. In both experiments feedback shapes subjects’ beliefs. With exogenously chosen feedback, transparent feedback tends to reduce beliefs when good examples are provided as feedback and tends to increase beliefs in when bad examples are provided as feedback compared to the respective non-transparent cases. Our results shed new light on the design of feedback provision in public goods settings.

Dishonest behavior significantly increases the cost of medical care provision. Upcoding of patients is a common form of fraud to attract higher reimbursements. Imposing audit mechanisms including fines to curtail upcoding is widely discussed among health care policy-makers. How audits and fines affect individual health care providers' behavior is empirically not well understood. To provide new evidence on fraudulent behavior in health care, we analyze the effect of a random audit including fines on individuals' honesty by means of a novel controlled behavioral experiment framed in a neonatal care context. Prevalent dishonest behavior declines significantly when audits and fines are introduced. The effect is driven by a reduction in upcoding when being detectable. Yet, upcoding increases when not being detectable as fraudulent. We find evidence that individual characteristics (gender, medical background, and integrity) are related to dishonest behavior. Policy implications are discussed.

Does it pay off for companies to disclose voluntary commitments to their customers? While voluntary commitments to enhance customers’ benefits became prevalent in many markets, systematic evidence on how customers (if at all) reward companies, which disclose such discretionary kindness, is still lacking. We analyze the consequences of endogenous disclosure of discretionary kindness in a novel experiment (N = 636). We model the decision situation in a bilateral reciprocity game with asymmetric information on the vol-untariness of kindness. Experimental data show that endogenously disclosing discretionary kindness significantly triggers rewards from customers and does not backfire. Findings are robust towards variations in costs of information and the level of customers’ benefits. Survey evidence from a vignette study support our behavioral findings.

Several studies show that social image concerns stimulate prosocial behavior. We study a setting in which there is uncertainty about which action is prosocial. Then, the quest for a better social image can potentially conflict with genuinely prosocial behavior. This conflict can induce “bad” behavior, where people lower both their own and others’ material payoffs to preserve a good image. This setting is relevant for various types of credence goods. For example, recommending an inexpensive treatment reduces the expert’s profits and may not satisfy the true needs of the client, but is generally good for the expert’s image (as it signals the lack of greed). We test experimentally if people start to act bad in order to look good. We find that people care about their social image, but social image concerns alone do not induce them to act bad. That is, without future interactions, social image concerns do not lead to bad behavior. However, with future interactions, where building up a good image has instrumental value (reputational concerns), we do find evidence of bad behavior in the short run to secure higher earnings in the long run.

We experimentally study the causal effect of being an immigrant or previously convicted on the hiring preferences and wage payments of employers. We find evidence for statistical discrimination against immigrants. Criminal offenders suffer from more severe and taste discrimination.

In a monetarily incentivized Dictator Game, we expected Dictators’ empathy toward the Recipients to cause more pro-social allocations. Empathy was experimentally induced via a commonly used perspective taking task. Dictators (N = 474) were instructed to split an endowment of 10€ between themselves and an unknown Recipient. They could split the money 8/2 (8€ for Dictator, 2€ for Recipient) or 5/5 (5€ each). Although the empathy manipulation successfully increased Dictators’ feelings of empathy toward the Recipients, Dictators’ decisions on how to split the money were not affected. We had ample statistical power (above 0.99) to detect a typical social psychology effect (corresponding to r around 0.20). Other possible determinants of generosity in the Dictator Game should be investigated.

With the advent of online subject pools, conducting experiments outside the laboratory has become more popular among the scientific community. Unlike the lab, online and field environments tend to be accompanied by a loss of control. In this article we introduce otree_tools, a concise yet powerful add-on for oTree (Chen et al., 2016). otree_tools provides novel ways by which to measure behaviors that are potentially important in the social sciences, such as attention, multitasking and effort. The software also features a novel method of tracking time through the identification of noise. We demonstrate the utility of otree_tools with the help of experimental evidence. The software substantially increases control of the environment. Moreover, the original metrics can be employed for innovative outcome variables, opening the avenue for new research opportunities.

A new class of intransitive objects – geometrical and mathematical constructions forming intransitive cycles A>B>C>A – are presented. In contrast to the famous intransitive dice, lotteries, etc., they show deterministic (not probabilistic) intransitive relations. The simplest ones visualize intransitivity that can be understood at a qualitative level and does not require quantitative reasoning. They can be used as manipulatives for learning intransitivity. Classification of the types of situations in which the transitivity axiom does and does not work is presented. Four levels of complexity of intransitivity are introduced, from simple combinatorial intransitivity to a "rhizomatic" one. A possible version of the main educational message for students in teaching and learning transitivity-intransitivity is presented.

The organization of collective action is extremely important for societies. A main reason is that many of the key problems societies face are public good problems. We present results from a series of laboratory experiments with large groups of up to 100 subjects. Our results demonstrate that large groups, in which the impact of an individual contribution ( MPCR ) is almost negligible, are able to provide a public good in the same way as small groups in which the impact of an individual contribution is much higher. Nevertheless, we find that small variations in MPCR in large groups have a strong effect on contributions. We develop a hypothesis concerning the interplay between MPCR and group size, which is based on the assumption that the salience of the advantages of mutual cooperation plays a decisive role. This hypothesis is successfully tested in a second series of experiments. Since Mancur Olson’s “Logic of collective action” it is a commonly held belief that in large groups the prospects of a successful organization of collective actions are rather bad. Our results, however, suggest that the chance to successfully organize collective action of large groups and to solve important public good problems is much higher than previously thought.

oTree, a popular platform for conducting behavioral experiments, exchanges data only as a participant exits or enters a web page. In many situations, however, information needs to be gathered and delivered instantaneously. This paper demonstrates a way to add real-time interactions to oTree and presents two ready-made apps: a double auction and a gift exchange with a real effort task. Many auction designs, including the double auction, use time constraints and carry out sales as soon as an ask and a bid are compatible. Instantaneous flow of information is thus a core requirement for programming these auctions in the first place. The real effort task measuring the number of correct answers within a time limit, on the other hand, benefits from the extra flexibility and security that Django Channels provides. Furthermore, real effort tasks are a simple starting point for building real-time interaction apps with oTree.

Recent studies have demonstrated that the right dorsolateral prefrontal cortex (rDLPFC) and the right temporoparietal junction (rTPJ) are causally involved in social norm compliance. Here, we tested the hypothesis that a third party's decision to punish norm violations depends on the activity of the entire rDLPFC/rTPJ network. We used transcranial direct current stimulation (tDCS) to independently or jointly modulate rTPJ and rDLPFC activity during the third-party dictator game. We found a significant effect of anodal tDCS of the rTPJ, which decreased the third-party punishment of moderately unfair splits. Joint stimulation of the rTPJ (by anodal tDCS) and rDLPFC (by cathodal tDCS) produced a marginal effect on third-party punishment.

In the commentary to Jens Mammen’s book A New Logical Foundation for Psychology (2017), three issues are discussed. The first one concerns possible interrelations of: (a) others’ irreplaceability and existential irretrievability rigorously proved by Mammen; and (b) morality and attitudes to the others. Lem’s criticism of Heidegger’s existential philosophy, which paradoxically ignores mass homicide, is discussed in the context of topology of being. Different attitudes to the other as irreplaceable and irretrievable (e.g., in case of apprehension and execution of a murderer) are analyzed. The second issue concerns the possibility of true duplicates of the same person. The paradox of copied complexity is introduced. The third issue concerns reductionism (including brain reductionism) and opportunities to deduce various phenomena of development (mental development, actual genesis of creative thinking, etc.) from the new logical foundation for psychology built by Mammen.