Deal with it
Cooperating with machines
Jacob Crandall et al.
Nature Communications, January 2018
Since Alan Turing envisioned artificial intelligence, technical progress has often been measured by the ability to defeat humans in zero-sum encounters (e.g., Chess, Poker, or Go). Less attention has been given to scenarios in which human–machine cooperation is beneficial but non-trivial, such as scenarios in which human and machine preferences are neither fully aligned nor fully in conflict. Cooperation does not require sheer computational power, but instead is facilitated by intuition, cultural norms, emotions, signals, and pre-evolved dispositions. Here, we develop an algorithm that combines a state-of-the-art reinforcement-learning algorithm with mechanisms for signaling. We show that this algorithm can cooperate with people and other algorithms at levels that rival human cooperation in a variety of two-player repeated stochastic games. These results indicate that general human–machine cooperation is achievable using a non-trivial, but ultimately simple, set of algorithmic mechanisms.
Pace of Life in Cities and the Emergence of Town Tweeters
Alexander Jones Gross, Dhiraj Murthy & Lav Varshney
SAGE Open, December 2017
Long-standing results in urban studies have shown correlation of population and population density to a city’s pace of life, empirically tested by examining whether individuals in bigger cities walk faster, spend less time buying stamps, or make greater numbers of telephone calls. Contemporary social media presents a new opportunity to test these hypotheses. This study examines whether users of the social media platform Twitter in larger and denser American cities tweet at a faster rate than their counterparts in smaller and sparser ones. Contrary to how telephony usage and productivity scale superlinearly with city population, the total volume of tweets in cities scales sublinearly. This is similar to the economies of scale in city infrastructures like gas stations. When looking at individuals, however, greater population density is associated with faster tweeting. The discrepancy between the ecological correlation and individual behavior is resolved by noting that larger cities have sublinear growth in the number of active Twitter users. This suggests that there is a more concentrated core of more active users that may serve an information broadcast function for larger cities, an emerging group of “town tweeters” as it were.
Group size in social-ecological systems
Marco Casari & Claudio Tagliapietra
Proceedings of the National Academy of Sciences, forthcoming
Cooperation becomes more difficult as a group becomes larger, but it is unclear where it will break down. Here, we study group size within well-functioning social-ecological systems. We consider centuries-old evidence from hundreds of communities in the Alps that harvested common property resources. Results show that the average group size remained remarkably stable over about six centuries, in contrast to a general increase in the regional population. The population more than doubled, but although single groups experienced fluctuations over time, the average group size remained stable. Ecological factors, such as managing forest instead of pasture land, played a minor role in determining group size. The evidence instead indicates that factors related to social interactions had a significant role in determining group size. We discuss possible interpretations of the findings based on constraints in individual cognition and obstacles in collective decision making.
When can shared attention increase affiliation? On the bonding effects of co-experienced belief affirmation
Parnia Haj-Mohamadi, Elizabeth Fles & Garriy Shteynberg
Journal of Experimental Social Psychology, March 2018, Pages 103-106
There is significant evidence suggesting that shared attention, or the perception of synchronous co-attention, can impact intrapersonal outcomes such as memory, affect, motivation, judgment, and behavior (Shteynberg, 2015). The interpersonal effects of shared attention, however, are not well-known. That is, we do not know if, and when, sharing attention changes feelings of closeness to the co-attendee. In an experiment with 447 participants, we examined whether sharing attention increased affiliation irrespective of the co-attended stimulus (the stimulus-independent hypothesis), or based on the co-attended stimulus (the stimulus-dependent hypothesis). The results substantiated the stimulus-dependent hypothesis, confirming that the effect of shared attention on affiliation depends on the stimulus under shared attention. More specifically, whereas belief-affirming messages under shared attention increased felt closeness to the co-attendee, belief-disaffirming and mixed messages under shared attention did not increase felt closeness to the co-attendee. In all, the findings suggest that sharing attention to belief affirming messages creates a unique situational context that enhances relational closeness.
The Advantage of Democratic Peer Punishment in Sustaining Cooperation within Groups
Stefan Pfattheicher, Robert Böhm & Rebekka Kesberg
Journal of Behavioral Decision Making, forthcoming
In social dilemma situations, individuals benefit from uncooperative behavior while exploiting resources of the collective. One prominent solution to prevent uncooperative behavior and to increase cooperation is to establish a system of costly peer punishment, that is, the possibility for every individual involved in the dilemma to impose costly punishment on interaction partners. However, recent research revealed that, in contrast to a situation without punishment, peer punishment is inefficient and maladaptive in the sense that the total payoff is reduced and punishment of cooperative individuals (i.e., antisocial punishment) is possible. In the present work, we propose that a system of democratic peer punishment, that is, direct and equal participation of each individual in the punishment decision-making process with punishment only executed when a majority has voted for its execution, can address the shortcomings of a peer punishment system. Using iterated public goods games, we show higher cooperation levels, higher total payoffs, and reduced executed punishment in the democratic compared with a peer punishment system. Moreover, we document that fairness perceptions, satisfaction, and interpersonal trust are increased in the democratic punishment system. Implications for how cooperation and democratic punishment systems may evolve are discussed.
Honest Abe or Doc Holliday? Bluffing in Bargaining
Gregory DeAngelo & Bryan McCannon
West Virginia University Working Paper, October 2017
We consider a bargaining environment where there is asymmetric information regarding whether the two players have common or conflicting preferences. If the cost of strategic communication is independent of the state, then signaling is not expected to be effective. If the uninformed agent believes, though, a (non-credible) signal has been sent, then the informed agents are incentivized to engage in deceptive bluffing. Alternatively, if bluffing is not too prevalent, honest communication can be worthwhile. We explore this theoretically and experimentally. We present a bargaining model where state-dependent mixed strategies arise as equilibria, i.e., bluffing. We then design an experimental game to assess the validity of the theoretical model’s predictions. We show that agents attempt to strategically transmit information even when (asymmetrically costly) signaling is not possible. Across rounds of the game honest, but not credible, signaling and bluffing co-move in that as the former becomes more prevalent so too does the latter. Furthermore, we document a contagion effect in the laboratory. Bluffing not only creates deadweight loss in a particular dyad, but leads the agent who was bluffed to engage in more bargaining conflict in future rounds against a new, randomly-selected opponent. Aggregate wealth is higher prior to the introduction of deception in the lab.