A variational-autoencoder approach to solve the hidden profile task in hybrid human-machine teams

Algorithmic agents, popularly known as bots, have been accused of spreading misinformation online and supporting fringe views. Collectives are vulnerable to hidden-profile environments, where task-relevant information is unevenly distributed across individuals. To do well in this task, information aggregation must equally weigh minority and majority views against simple but inefficient majority-based decisions. In an experimental design, human volunteers working in teams of 10 were asked to solve a hidden-profile prediction task. We trained a variational auto-encoder (VAE) to learn people's hidden information distribution by observing how people's judgements correlated over time. A bot was designed to sample responses from the VAE latent embedding to selectively support opinions proportionally to their under-representation in the team. We show that the presence of a single bot (representing 10\% of team members) can significantly increase the polarization between minority and majority opinions by making minority opinions less prone to social influence. Although the effects on hybrid team performance were small, the bot presence significantly influenced opinion dynamics and individual accuracy. These findings show that self-supervised machine learning techniques can be used to design algorithms that can sway opinion dynamics and group outcomes.

A variational-autoencoder approach to solve the hidden profile task in hybrid human-machine teams

Online, social media bots have been accused to spread misinformation and support extreme or minority-held opinions. However, bots in hybrid human-machine teams can also be designed to improve team performance. In this paper, we study the effect of a single minority-supporting bot in hybrid teams in a carefully controlled experiment. People working in teams of 10 were asked to solve a hidden-profile prediction task, where task-relevant information was scattered unequally across team members. To do well in this task, pieces of information shared by the minority and the majority of players should be integrated. Simple majority-based decisions are not enough to perform well as information held by minority players is also valuable. We used a variational auto-encoder to train a bot to learn people's information distribution by observing how people's judgements correlated over time. After training, a bot was designed to increase team performance by selectively supporting opinions proportionally to their under-representation in the team. We show that the presence of a single bot (representing 10\% of team members) can significantly increase the polarization between minority and majority opinions by making minority opinions less prone to social influence. Although the effects on hybrid team performance were negligible, the bot presence significantly influenced team opinion dynamics. These findings show that unsupervised learning can be used to program bots that can improve team performance.

Sharing Sensitive Information: A Hidden Profile Employee Selection Exercise

Employee selection and group decision-making skills are critical for ensuring hiring is valid, meets organizational goals, and considers ethical and legal limitations. This exercise has participants role-play members of a search committee reviewing job finalists using shared and unique information. A novel twist to traditional hidden-profile exercises is introduced by including unique information inappropriate for employment decisions (e.g., health information, an old misdemeanor charge). By uncovering unshared details and deciding whether to discuss potentially biasing information, learners practice group decision making and consider legal issues. While exploring professional guidelines and best practices, the exercise acknowledges that managers occasionally know sensitive or potentially biasing information. Although primarily an human resource activity, the exercise includes a traditional hidden-profile variant with organizational behavior learning goals. Both variants are appropriate for learners across the organizational spectrum. A teaching note for adapting the in-person exercise for synchronous or asynchronous online delivery gives detailed instructions for popular learning management systems.

Equal Performance, Different Grade: Women’s Performance in Discussion Perceived Worse Than Men’s

We investigate how men and women are evaluated in group discussions. In five studies ( N = 761) using a variant of a Hidden Profile Task, we find that, when experimentally and/or statistically controlling for actual gender differences in behavior, the female performance in a group discussion is devalued in comparison to male performance. This was observed for fellow group members (Study 1) and outside observers (Studies 2–5), in both primarily student (Studies 1, 4, and 5) and mixed samples (Studies 2 and 3), for different measures of performance (perceived helpfulness of the contribution, for work-related competence), across different discussion formats (preformulated chat messages, open chat), and when controlling for the number of female group members (Study 5). In contrast to our hypothesis, we did not find a moderating effect of selection procedure in that women were devalued to a similar degree in both situations with a women’s quota and without.

Quality of group decisions by board members: a hidden-profile experiment

PurposeReaching decisions in a deliberative manner is of utmost importance for boards, as their decision-making impacts entire organisations. The current study aims to investigate (1) the quality of group decisions made by board members, (2) their confidence in, satisfaction with, and reflection on the decision-making, and (3) the effect of two discussion procedures on objective decision quality and subjective evaluations of the decision-making.Design/methodology/approachBoard members of various Dutch non-profit organisations (N = 141) participated in a group decision-making task and a brief questionnaire. According to the hidden-profile paradigm, information was asymmetrically distributed among group members and should have been pooled to reach the objectively best decision. Half of the groups received one of two discussion procedures (i.e. advocacy decision or decisional balance sheet), while the other half received none.FindingsOnly a fifth of the groups successfully chose the best decision alternative. The initial majority preference strongly influenced the decision, which indicates that discussion was irrelevant to the outcome. Nevertheless, board members were satisfied with their decision-making. Using a discussion procedure enhanced participants' perception that they adequately weighed the pros and cons, but did not improve objective decision quality or other aspects of the subjective evaluation. These findings suggest that board members are unaware of their biased decision-making, which might hinder improvement.Originality/valueRather than using student samples, this study was the first to have board members participating in a hidden-profile task.

Mental simulation and the individual preference effect

PurposeThe “Individual Preference Effect” (IPE: Faulmüller et al., 2010; Greitemeyer and Schulz-Hardt, 2003; Greitemeyer et al., 2003), a form of confirmation bias, is an important barrier to achieving improved group decision-making outcomes in hidden profile tasks. Group members remain committed to their individual preferences and are unable to disconfirm their initial suboptimal selection decisions, even when presented with full information enabling them to correct them, and even if the accompanying group processes are perfectly conducted. This paper examines whether a mental simulation can overcome the IPE.Design/methodology/approachTwo experimental studies examine the effect of a mental simulation intervention in attenuating the IPE and improving decision quality in an online individual hidden profile task.FindingsIndividuals undertaking a mental simulation achieved higher decision quality than those in a control condition and experienced a greater reduction in confidence in the suboptimal solution.Research limitations/implicationsResults suggest a role for mental simulation in overcoming the IPE. The test environment is an online individual decision-making task, and broader application to group decision-making is not tested.Practical implicationsSince mental simulation is something we all do, it should easily generalise to an organisational setting to improve decision outcomes.Originality/valueTo the authors' knowledge, no study has examined whether mental simulation can attenuate the IPE.

Collective Choice, Collaboration, and Communication

This article reviews recent empirical research on collective choice and collaborative problem solving. Much of the collective choice research focuses on hidden profiles. A hidden profile exists when group members individually have information favoring suboptimal choices but the group collectively has information favoring an optimal choice. Groups are notoriously bad at discovering optimal choices when information is distributed to create a hidden profile. Reviewed work identifies informational structures, individual processing biases, and social motivations that inhibit and facilitate the discovery of hidden profiles. The review of collaborative problem-solving research is framed by Larson's concept of synergy. Synergy refers to performance gains that are attributable to collaboration. Recent research has addressed factors that result in groups performing as well as their best member (weak synergy) and better than their best member (strong synergy). Communication dynamics underlying both collective choice and collaborative problem solving are discussed.