Repeated Interactions Research Articles

Single-molecule electrical spectroscopy of organocatalysis

•Universal single-molecule electrical spectroscopy to monitor complex reactions •Novel reaction dynamics and new intermediates discovered by correlation analysis •New thermodynamic/kinetic model offering quantitative analysis on a molecular basis Conventional mechanistic studies based on detection of species concentration reveal the basic laws of matter conversion, but inherently display ensemble averages. Single-molecule detection provides a solution and direct visualization of reaction pathway rather than only detection of species. Here, we establish an in situ label-free electrical detection platform and report universal single-molecule electrical spectroscopy (SMES) of organocatalysis to decipher the intrinsic mechanism of a benzoin reaction. This SMES is capable of monitoring each event in a simple (homo-condensation) or complex (cross-condensation) system. Both experimental and theoretical results consistently revealed accuracy of the spectroscopic attribution and proved its reliability. Furthermore, the correlation of adjacent events in the spectroscopy indicated the local repeated interaction between a catalyst and a substrate. Through a new thermodynamic and kinetic model, the inventory of substrates around the single-molecule catalyst can be reversely deduced, which enables quantitative analysis of the target single molecule. Conventional mechanistic studies based on detection of species concentration reveal the basic laws of matter conversion, but inherently display ensemble averages. Single-molecule detection provides a solution and direct visualization of reaction pathway rather than only detection of species. Here, we establish an in situ label-free electrical detection platform and report universal single-molecule electrical spectroscopy (SMES) of organocatalysis to decipher the intrinsic mechanism of a benzoin reaction. This SMES is capable of monitoring each event in a simple (homo-condensation) or complex (cross-condensation) system. Both experimental and theoretical results consistently revealed accuracy of the spectroscopic attribution and proved its reliability. Furthermore, the correlation of adjacent events in the spectroscopy indicated the local repeated interaction between a catalyst and a substrate. Through a new thermodynamic and kinetic model, the inventory of substrates around the single-molecule catalyst can be reversely deduced, which enables quantitative analysis of the target single molecule.

When network bridges foster consensus. Bounded confidence models in networked societies

In this work we present novel results to the problem of the Hegselmann-Krause dynamics in networks obtained by an extensive study of the behavior of the standard order parameter sensitive to the onset of consensus: the normalized size of the giant cluster. This order parameter reveals the non trivial effect of the network topology on the steady states of the dynamics, overlooked by previous works, which concentrated on the onset of unanimity, and allows to detect regions of polarization between the fragmented and the consensus phases. While the previous results on unanimity are confirmed, the consensus threshold shifts in the opposite direction compared to the threshold for unanimity. A detailed finite size scaling analysis shows that, in general, consensus is easier to obtain in networks than in mixed populations. At a difference with previous studies, we show that the network topology is relevant beyond the finitness of the average degree with increasing system size. In particular, in pure random networks (either uniform random graphs or scale free networks), the consensus threshold seems to vanish in the thermodynamic limit. A detailed analysis of the time evolution of the dynamics reveals the role of bridges in the network, which allow for the interaction between agents belonging to clusters of very different opinions, after several repeated interaction steps. These bridges are at the origin of the shift of the confidence threshold to lower values in networks as compared to lattices or the mixed population.

When leading by example leads to less corrupt collaboration

We contribute to the pressing question of how organizational design influences corporate wrongdoing by studying different decision structures — simultaneous vs. sequential — in experimental coordination games. Participants can report private information honestly, or lie to increase their own, as well as the group’s, payoff. In simultaneous decision structures, all group members report at the same time, without information about the reports of others, whereas in sequential decision structures there is a first mover who decides first. We find that the presence of a first mover decreases dishonesty levels in repeated interactions (but not in one-shot settings). We argue that this effect is primarily driven by image concerns of decision leaders.

Examining Social Cognition with Embodied Robots: Does Prior Experience with a Robot Impact Feedback-associated Learning in a Gambling Task?

Social agents rely on the ability to use feedback to learn and modify their behavior. The extent to which this happens in social contexts depends on motivational, cognitive and/or affective parameters. For instance, feedback-associated learning occurs at different rates when the outcome of an action (e.g., winning or losing in a gambling task) affects oneself (“Self”) versus another human (“Other”). Here, we examine whether similar context effects on feedback-associated learning can also be observed when the “other” is a social robot (here: Cozmo). We additionally examine whether a “hybrid” version of the gambling paradigm, where participants are free to engage in a dynamic interaction with a robot, then move to a controlled screen-based experiment can be used to examine social cognition in human-robot interaction. This hybrid method is an alternative to current designs where researchers examine the effect of the interaction on social cognition during the interaction with the robot. For that purpose, three groups of participants (n total = 60) interacted with Cozmo over different time periods (no interaction vs. a single 20 minute interaction in the lab vs. daily 20 minute interactions over five consecutive days at home) before performing the gambling task in the lab. The results indicate that prior interactions impact the degree to which participants benefit from feedback during the gambling task, with overall worse learning immediately after short-term interactions with the robot and better learning in the “Self” versus “Other” condition after repeated interactions with the robot. These results indicate that “hybrid” paradigms are a suitable option to investigate social cognition in human-robot interaction when a fully dynamic implementation (i.e., interaction and measurement dynamic) is not feasible.

Neighbourhood places, collective efficacy and crime: A longitudinal perspective

Neighbourhood places like shops, cafes and parks support a variety of social interactions ranging from the ephemeral to the intimate. Repeated interactions at neighbourhood places over time lay the foundation for the development of social cohesion and collective efficacy. In this study, we examine the proposition that changes in the presence or arrangement of neighbourhood places can destabilise social cohesion and collective efficacy, which has implications for crime. Using spatially integrated crime, social survey and parcel-level land-use classification data, we estimate mixed effects panel models predicting changes in theft and nuisance crimes across 147 Australian neighbourhoods. The findings are consistent with neighbourhood social control and crime opportunity theories. Neighbourhood development – indicated by fewer vacant properties and fewer industrial and agricultural sites – is associated with higher collective efficacy and less crime over time. Conversely, introducing more restaurants, transit stations and cinemas is associated with higher theft and nuisance over time regardless of neighbourhood collective efficacy. We argue that the addition of socially conducive places can leave neighbourhoods vulnerable to crime until new patterns of sociability emerge and collective efficacy develops.

The political economy of industrial pollution control: environmental regulation in Swedish industry for five decades

The paper analyzes the prerequisites for a regulatory-driven transition toward radically lower air and water pollution in industry. This is achieved in the empirical context of the Swedish mining and metals industry, and by investigating the environmental licensing processes during two regulatory systems. The paper derives an analytical framework that explores under what circumstances such licensing processes can result in radical emissions reductions without seriously jeopardizing the competitiveness of the industry. Archived material covering six environmental licensing processes, three during each system, is used to illustrate the various design and implementation issues. The results suggest that regulatory-driven green transitions benefit from trust-based bargaining procedures in which companies are involved in repeated interactions with regulatory authorities, and which extended probation periods permit tests of novel abatement technologies (including innovation). The findings also illustrate the importance of abstaining from simplified normative notions about policy instrument choice (e.g. taxes versus standards).

Human-Aware Reinforcement Learning for Fault Recovery Using Contextual Gaussian Processes

This work addresses the iterated nonstationary assistant selection problem, in which over the course of repeated interactions on a mission, an autonomous robot experiencing a fault must select a single human from among a group of assistants to restore it to operation. The assistants in our problem have a level of performance that changes as a function of their experience solving the problem. Our approach uses reinforcement learning via a multi-arm bandit formulation to learn about the capabilities of each potential human assistant and decide which human to task. This study, which is built on our past work, evaluates the potential for a Gaussian-process-based machine learning method to effectively model the complex dynamics associated with human learning and forgetting. Application of our method in simulation shows that our method is capable of tracking performance of human-like dynamics for learning and forgetting. Using a novel selection policy called the proficiency window, it is shown that our technique can outperform baseline selection strategies while providing guarantees on human use. Our work offers an effective potential alternative to dedicated human supervisors, with application to any human–robot system where a set of humans is responsible for overseeing autonomous robot operations.

I Can See It in Your Eyes: Gaze as an Implicit Cue of Uncanniness and Task Performance in Repeated Interactions With Robots.

Over the past years, extensive research has been dedicated to developing robust platforms and data-driven dialog models to support long-term human-robot interactions. However, little is known about how people's perception of robots and engagement with them develop over time and how these can be accurately assessed through implicit and continuous measurement techniques. In this paper, we explore this by involving participants in three interaction sessions with multiple days of zero exposure in between. Each session consists of a joint task with a robot as well as two short social chats with it before and after the task. We measure participants' gaze patterns with a wearable eye-tracker and gauge their perception of the robot and engagement with it and the joint task using questionnaires. Results disclose that aversion of gaze in a social chat is an indicator of a robot's uncanniness and that the more people gaze at the robot in a joint task, the worse they perform. In contrast with most HRI literature, our results show that gaze toward an object of shared attention, rather than gaze toward a robotic partner, is the most meaningful predictor of engagement in a joint task. Furthermore, the analyses of gaze patterns in repeated interactions disclose that people's mutual gaze in a social chat develops congruently with their perceptions of the robot over time. These are key findings for the HRI community as they entail that gaze behavior can be used as an implicit measure of people's perception of robots in a social chat and of their engagement and task performance in a joint task.

Collision Models Can Efficiently Simulate Any Multipartite Markovian Quantum Dynamics.

We introduce the multipartite collision model, defined in terms of elementary interactions between subsystems and ancillas, and show that it can simulate the Markovian dynamics of any multipartite open quantum system. We develop a method to estimate an analytical error bound for any repeated interactions model, and we use it to prove that the error of our scheme displays an optimal scaling. Finally, we provide a simple decomposition of the multipartite collision model into elementary quantum gates, and show that it is efficiently simulable on a quantum computer according to the dissipative quantum Church-Turing theorem, i.e., it requires a polynomial number of resources.

What Situation Is This? Shared Frames and Collective Performance

We study agents who distill the complex world around them using cognitive frames. We assume that agents share the same frame and analyze how the frame affects their collective performance. In one-shot and repeated interactions, the frame causes agents to be either better or worse off than if they could perceive the environment in full detail: it creates a fog of cooperation or a fog of conflict. In repeated interactions, the frame is as important as agents’ patience in determining the outcome: for a fixed discount factor, when all agents choose what they perceive as their best play, there remain significant performance differences induced by different frames. A low-performing team conducting a site visit to observe a high-performing team will be mystified, sometimes observing different actions than they expected or being given unexpected reasons for the actions they expected. Finally, we distinguish between incremental versus radical changes in frames, and we develop a model of category formation to analyze challenges faced by a leader who seeks to improve the agents’ collective performance.

Arbitration supports reciprocity when there are frequent perception errors.

Reciprocity is undermined by perception errors, mistakes that cause disagreement between interacting individuals about past behaviour. Strategies such as win-stay-lose-shift and generous tit-for-tat can re-establish cooperation following a perception error, but only when errors arise infrequently. We introduce arbitration tit-for-tat (ATFT), a strategy that uses third-party arbitration to align players' beliefs about what transpired when they disagree. We show that, when arbitration is moderately accurate, ATFT is a strong subgame-perfect equilibrium and is evolutionarily stable against a range of strategies that defect, cooperate, ignore arbitration or invoke arbitration unnecessarily. ATFT can persist when perception errors are frequent, arbitration is costly or arbitration is biased. The need for third parties to resolve perception errors could explain why reciprocity is rare in other animals despite opportunities for repeated interactions and why human reciprocity is embedded within culturally transmitted moral norms in which community monitoring plays a role.

Experimental cheap talk games: strategic complementarity and coordination

This paper investigates experimentally the effects of communication in distinct games with complete information. We design four games resulting from the interaction between two incentive elements: strategic complementarity and coordination. These incentive elements allow to analyse the use of cheap talk as an efficiency-enhancing and coordinating device. We implement a restricted communication protocol (one-sided, optional, and closed-form) in repeated settings with fixed partners. Our findings provide robust evidence about how cheap talk interacts with incentives to explain strategic behaviour in a dynamic way. As expected, cheap talk increases efficiency under complementarity incentives, and the coordination level under coordination incentives. As novelty, the use of limited communication in repeated interactions has led to identify specific time-varying message profiles as the most effective messages in the coordination games. While the content of messages is explained by the complementarity incentives, faithfulness to credible messages is determined by the coordination incentives.

On existence of private unemployment insurance with advance information on future job losses

We study the existence of a profitable unemployment insurance market in a dynamic economy with adverse selection rooting in information on future job losses. The new feature of the model is that the insurer and workers interact repeatedly. Repeated interactions make it possible to threaten workers with exclusion from future insurance benefits after a default on insurance premia. With exclusion, not only the insurance against the fundamental risk, but also against future bad news about job losses matters. In contrast to conventional wisdom, we find that private unemployment insurance in the US can be profitable for a relatively short exclusion length of one year. To stimulate the emergence of a private unemployment insurance market, policy makers can facilitate the creation of a registry that archives past defaults on insurance premia.

The Collusive Efficacy of Competition Clauses in Bertrand Markets with Capacity-Constrained Retailers

We study the collusive efficacy of competition clauses (CC) such as the meeting competition clause (MCC) and the beating competition clauses (BCC) in a general framework. In contrast to previous theoretical studies, we allow for repeated interaction among the retailers and heterogeneity in their sales capacities. Besides that, the selection of the form of the CC is endogeneized. The retailers choose among a wide range of CC types - including the conventional ones such as the MCC and the BCCs with lump sum refunds. Several common statements about the collusive (in)efficacy of CCs cannot be upheld in our framework. We show that in the absence of hassle costs, MCCs might induce collusion in homogeneous markets even if they are adopted only by few retailers. If hassle and implementation costs are mild, collusion can be enforced by BCCs with lump sum refunds. Remarkably, these fundings hold for any reasonable rationing rule. However, a complete specification of all collusive CCs is in general impossible without any further reference to the underlying rationing rule.

Reinforcement learning decoders for fault-tolerant quantum computation

Topological error correcting codes, and particularly the surface code, currently provide the most feasible road-map towards large-scale fault-tolerant quantum computation. As such, obtaining fast and flexible decoding algorithms for these codes, within the experimentally realistic and challenging context of faulty syndrome measurements, without requiring any final read-out of the physical qubits, is of critical importance. In this work, we show that the problem of decoding such codes can be naturally reformulated as a process of repeated interactions between a decoding agent and a code environment, to which the machinery of reinforcement learning can be applied to obtain decoding agents. While in principle this framework can be instantiated with environments modelling circuit level noise, we take a first step towards this goal by using deepQ learning to obtain decoding agents for a variety of simplified phenomenological noise models, which yield faulty syndrome measurements without including the propagation of errors which arise in full circuit level noise models.

Incomplete Political Contracts with Secret Ballots: Reciprocity as a Force to Enforce Sustainable Clientelistic Relationships

Abstract Clientelism is frequently observed in our societies. Various mechanisms that help sustain incomplete political contracts (e.g., monitoring and punishment) have been studied in the literature to date. However, do such contracts emerge in elections with secret ballots when the interactions are one-shot? How does repetition affect the evolution of incomplete political contracts? Using an incentivized experiment, this article finds that even during one-shot interactions where monitoring is not possible, candidates form incomplete contracts through vote buying and promise-making. The candidates’ clientelistic behaviors are heterogeneous: some target swing voters, whereas others offer the most to loyal voters, or even opposition voters. These tactics distort voting behaviors as well as election outcomes. Repeated interactions significantly magnify candidates’ offers and deepen clientelistic relationships. These results underscore the possibility that clientelism evolves due to people’s strategic behaviors and interdependent preferences, without relying on alternative mechanisms (JEL C92, D72).

Paper-Rock-Scissors: an exploration of the dynamics of players’ strategies

This research contributes to the understanding of dynamic decision making behavior in adversarial repeated interactions. Using a well-known competitive game, Rock-Paper-Scissors in a two-player experiment, we collected data of repeated play in pairs over many trials. We design a payoff matrix that allows us to distinguish the optimal (Nash) behavior from random behavior. Our analyses indicate that participants do not play in agreement with Nash or random. We also do not find evidence of the cyclic behavior suggested in the literature. Interestingly, human behavior is very heterogeneous. While some players follow the common “Win-Stay/Lose-Shift” heuristic, many others also follow a "Win-Shift/Lose-Stay” heuristic. We summarize our conclusions for the study of the dynamics of behavior in adversarial situations.

Understanding the relation between repeat developer interactions and bug resolution times in large open source ecosystems: A multisystem study

AbstractLarge‐scale software systems are being increasingly built by distributed teams of developers who interact across geographies and time zones. Ensuring smooth knowledge transfer and the percolation of skills within and across such teams remain key challenges for organizations. Towards addressing this challenge, organizations often grapple with questions around whether and how repeat collaborations between members of a team relate to outcomes of important activities. In the context of this paper, the word ‘repeat interaction’ does not imply a greater number of interactions; it refers to repeat interaction between a pair of developers who have collaborated before. In this paper, we empirically examine such a question using real‐world data from three diverse development ecosystems, collectively involving 400,000+ units of work and 600,000+ comments exchanged between numerous developers. Our statistical models consistently establish a counter‐intuitive relation between repeat developer interaction and bug resolution times. Our experimental results show that more instances of repeat developer interactions over bug fixing are associated with more time taken for the bugs to be fixed. Given the expanse and variety of the underlying data, our results offer an unexpected set of insights on a key dynamic of collaboration in software development ecosystems. We discuss how these insights can influence the practice of large‐scale software development at individual, team and organizational levels.

Particle Swarm Optimization on Deep Reinforcement Learning for Detecting Social Spam Bots and Spam-Influential Users in Twitter Network

In online social networks (OSNs), detection of malicious social bots is an important research challenge to provide legitimacy of user profiles and trustworthy service ratings. Further, spam-influential users must be minimal to control the fake information-spread in OSNs. Learning from example patterns using supervised learning may not provide accurate results in cases where existing data items are biased and bot behavior dynamically changes over a period of time. Moreover, deep reinforcement learning provides improved learning by repeated interactions with the environment. However, a typical deep reinforcement leaning algorithm converges slower to find an optimal sequence of actions to reach out a goal state. In this article, we design a particle swarm optimization (PSO) based deep Q-learning algorithm for detecting social spam bots by integrating PSO with Q-value function. In addition, spam-influential users are identified using the proposed spam influence minimization model and it helps in restricting the flow of illegitimate tweets in Twitter network. Further, an influential community detection algorithm has been proposed to reduce the spreading of spam content through influential communities in Twitter network. Experimental results illustrate the efficacy of our proposed algorithms by considering two Twitter datasets and performance metrics such as precision, recall, and modularity.

Social Proximity and the Evolution of Norm Compliance

We study how individuals' compliance with norms of pro-social behavior is influenced by other actors' compliance in a novel, dynamic, and non-strategic experimental setting. We are particularly interested in the role that social proximity among peers plays in eroding or upholding norm compliance. Our results suggest that social proximity is crucial. In settings without known proximity, norm compliance erodes swiftly because participants only conform to observed norm violations of their peers while ignoring norm compliance. With known social proximity, participants conform to both types of observed behaviors, thus halting the erosion of norm compliance. Our findings stress the importance of the broader social context for norm compliance and show that, even in the absence of social sanctions, compliance can be sustained in repeated interactions, provided there is group identification, as is the case in many social encounters in natural and online environments.