Conversational technology and reactions to withheld information

In this chapter, we examine connections between COVID-19 and climate, the two major crises we have been living through over the past few years. How has the former affected the world's collective. response to the latter? What lessons have people learned about climate change from COVID-19, and what lessons should we (and policymakers) learn?

Two distinct literatures have shown that people are averse to being rejected when requesting help, and to rejecting others’ requests for help; both parties prefer situations in which no ask was made to those in which an ask was made but rejected. We propose a game-theoretic framework that offers a parsimonious explanation for both phenomena and further extends them. In this framework, people who need help make inferences and care about what an offer of help, compliance with a help request, or rejection of a request signals about the potential help-giver’s concern for them or the relationship. Likewise, potential help-givers care about the inferences potential help-receivers make, as they want to appear caring. We propose that both the aversion to being rejected after an ask and the aversion to rejecting others after an ask can be explained by the fact that rejection provides a negative signal about how much the would-be helper cares about or values the person in need. The framework further predicts that the same mechanism leads to an aversion to asks even when help is provided. That is, holding constant whether help is provided, both parties incur a psychological cost whenever there is an ask. Two studies, one involving recollections of help-related experiences, provide empirical support for the framework. By bringing together disparate literatures in help-seeking and-giving, we uncover common psychological features underlying these economically and socially important behaviors, generate novel insights into how they can be encouraged, and draw connections to behaviors in related domains.Abstract: 250 words

Available at SSRN 4759434

In a recent paper, Chater and Loewenstein (2022) argue that behavioral scientists have been testing and advocating individualistic (i-frame) solutions to policy problems that have systemic (s-frame) causes and require systemic solutions. Here, we consider the implications of adopting an s-frame approach for research. We argue that an s-frame approach will involve addressing different types of questions, which will, in turn, require a different toolbox of research methods.

arXiv preprint arXiv:2402.13272

Existing approaches to Theory of Mind (ToM) in Artificial Intelligence (AI) overemphasize prompted, or cue-based, ToM, which may limit our collective ability to develop Artificial Social Intelligence (ASI). Drawing from research in computer science, cognitive science, and related disciplines, we contrast prompted ToM with what we call spontaneous ToM -- reasoning about others' mental states that is grounded in unintentional, possibly uncontrollable cognitive functions. We argue for a principled approach to studying and developing AI ToM and suggest that a robust, or general, ASI will respond to prompts \textit{and} spontaneously engage in social reasoning.

Journal of Experimental Social Psychology

Research on gender differences in (self-)disclosure has produced mixed results, and, where differences have emerged, they may be an artifact of the measures employed. The present paper explores whether gender – defined as self-identified membership in one's sociocultural group – can indeed account for differences in the desire and propensity to divulge information to others. We additionally identify a possible moderator for such differences. In three studies employing two distinct research approaches – a free recall task for the extreme desire to disclose (Study 1, N = 195) and scaled responses to scenarios that manipulate valence experimentally in an exploratory study (Study 2, N = 547) and a preregistered replication (Study 3, N = 405) – we provide evidence of a robust interaction between gender and information valence. Male participants appear similar to female participants in their desire and likelihood to …

arXiv preprint arXiv:2402.13273

We explore the use of aggregative crowdsourced forecasting (ACF) as a mechanism to help operationalize ``collective intelligence'' of human-machine teams for coordinated actions. We adopt the definition for Collective Intelligence as: ``A property of groups that emerges from synergies among data-information-knowledge, software-hardware, and individuals (those with new insights as well as recognized authorities) that enables just-in-time knowledge for better decisions than these three elements acting alone.'' Collective Intelligence emerges from new ways of connecting humans and AI to enable decision-advantage, in part by creating and leveraging additional sources of information that might otherwise not be included. Aggregative crowdsourced forecasting (ACF) is a recent key advancement towards Collective Intelligence wherein predictions (X\% probability that Y will happen) and rationales (why I believe it is this probability that X will happen) are elicited independently from a diverse crowd, aggregated, and then used to inform higher-level decision-making. This research asks whether ACF, as a key way to enable Operational Collective Intelligence, could be brought to bear on operational scenarios (i.e., sequences of events with defined agents, components, and interactions) and decision-making, and considers whether such a capability could provide novel operational capabilities to enable new forms of decision-advantage.

Available at SSRN

Policy challenges can typically be addressed both through systemic changes (eg, taxes and mandates) and by encouraging individual behavior change. In this paper, we propose that, while in principle complementary, systemic and individual perspectives can compete for the limited attention of people and policymakers. Thus, directing policies in one of these two ways can distract the public’s attention from the other—an “attentional opportunity cost.” In two pre-registered experiments (n= 1,800) covering three high-stakes domains (climate change, retirement savings, and public health), we show that when people learn about policies targeting individual behavior (such as awareness campaigns), they are more likely to themselves propose policies that target individual behavior, and to hold individuals rather than organizational actors responsible for solving the problem, than are people who learned about systemic policies (such as taxes and mandates, Study 1). This shift in attribution of responsibility has behavioral consequences: people exposed to individual interventions are more likely to donate to an organization that educates individuals rather than one seeking to effect systemic reforms (Study 2). Policies targeting individual behavior may, therefore, have the unintended consequence of redirecting attention and attributions of responsibility away from systemic change to individual behavior.

Many choices are complex. These choices entail consideration of multiple variables characterized by non-linear interactions. Prior research into the complex choice process of people neglects two categories of major choice determinants: heuristics and biases. This research takes on the challenge of integrating such behavior determinants into the study of complex choices. We do so using a new experimental paradigm in which participants tune on-screen dials that represent choice variables. Through their tuning efforts, participants can discover information about the choice space. Specifically, this new paradigm allows us to document how anchoring and framing effects impact the complex choice process. We find that loss framing incites participants to expend more effort examining choice variables before selecting an option. An aspirational anchor, on the other hand, was correlated with decreased search effort by participants. Interestingly, the two did not share an interaction. Additionally, when participants did not have an aspirational anchor, the data suggest they relied on prior experience for an anchor value.

Successful human-machine teaming often hinges on the ability of eXplainable Artificial Intelligence (XAI) to make an agent’s reasoning transparent to human teammates. Doing so requires that the agent navigate a tradeoff between revealing its reasoning to those teammates without overwhelming them with too much information. This challenge is amplified when a person is teamed with multiple agents. This amplification is not simply linear, due to the increase from 1 to N agents’ worth of reasoning content, but also due to the interdependency among the agents’ reasoning that must be made transparent as well. In this work, we examine the challenges in conveying this interdependency to people teaming with multiple agents. We also propose alternate domain-independent strategies for a team of simulated robots to generate messages about their reasoning to be conveyed to a human teammate. We illustrate these …

Management Science

Using data from an online brokerage, we examine the role of investor logins in trading behavior. We find that a new reference point is created when an investor logs in and views the investor’s portfolio. We observe this as a disposition effect on returns since last login in addition to the traditional disposition effect on returns since purchase. Further, these reference points produce a strong interaction such that even a small loss since last login nullifies the positive effect of a gain since purchase. This interaction follows if investors select the higher, more aspirational price as a reference point.This paper was accepted by Yuval Rottenstreich, behavioral economics and decision analysis.Funding: This work was supported by the Economic and Social Research Council [Grants ES/K002201/1, ES/N018192/1, ES/P008976/1, and ES/V004867/1] and the Leverhulme Trust [Grant RP2012-V-022].Supplemental Material: The …

Behavioral & Brain Sciences

Our target article distinguishes between policy approaches that seek to address societal problems through intervention at the level of the individual (adopting the" i-frame") and those that seek to change the system within which those individuals live (adopting the" s-frame"). We stress also that a long-standing tactic of corporations opposing systemic change is to promote the i-frame perspective, presumably hoping that i-frame interventions will be largely ineffective and more importantly will be seen by the public and some policy makers as a genuine alternative to systemic change. We worry that the i-frame focus of much of behavioral science has inadvertently reinforced this unhelpful focus on the individual. In this response to commentators, we identify common themes, build on the many constructive suggestions to extend our approach, and reply to concerns. We argue, along with several commentators, that a key …

Available at SSRN 4368304

Attention is a pivotal resource in the modern economy and plays an increasingly prominent role in economic analysis. We summarize research on attention from both psychology and economics, placing a particular emphasis on its capacity to explain numerous documented violations of classical economic theory. We also propose promising new directions for future research, including attention-based utility, the recent proliferation of attentional externalities introduced by digital technology, the potential for artificial intelligence to compete with human attention, and the significant role that boredom, curiosity, and other motivational states play in determining how people allocate attention.

Current Directions in Psychological Science

The literature on privacy-related behaviors and preferences often frames disclosure as strategic—the result of a weighing of costs and benefits and a pursuit of instrumental benefits rather than as a goal in and of itself. In the present article, we summarize evidence supporting the view that disclosure can exhibit drive-like qualities and that this “drive to disclose” can, at times, overwhelm the motive to maintain privacy. We discuss implications of this perspective, highlighting ways in which recognizing the existence of a drive to disclose can inform privacy research and policy making.

This book constitutes the proceedings of the First International Symposium, ToM for Teams 2021, held in Washington, DC, USA, during November 4–6, 2021, Each chapter in this section tackles a different aspect of AI representing the thoughts and beliefs of human agents. The work presented herein represents our collective efforts to better understand ToM, develop AI with ToM capabilities (ASI), and study how to integrate such systems into human teams.

An implicit expectation of asking users to rate agents, such as an AI decision-aid, is that they will use only relevant information—ask them about an agent’s benevolence, and they should consider whether or not it was kind. Behavioral science, however, suggests that people sometimes use irrelevant information. We identify an instance of this phenomenon, where users who experience better outcomes in a human-agent interaction systematically rated the agent as having better abilities, being more benevolent, and exhibiting greater integrity in a post hoc assessment than users who experienced worse outcomes—which were the result of their own behavior—with the same agent. Our analyses suggest the need for augmentation of models so they account for such biased perceptions as well as mechanisms so that agents can detect and even actively work to correct this and similar biases of users.

Journal of Behavioral Decision Making

Many situations involving exploration, such as businesses expanding into new products or locations, expose the explorer to the potential for subjective losses. How does the potential to experience losses during the course of a search affect individuals' appetite for exploration? In three incentivized studies, we manipulate search outcomes by presenting participants either with a gain‐only environment or a gain‐loss environment. The two environments offer objectively identical incentives for exploration: Using a framing manipulation, we decrease gain‐loss payoffs and provide participants an initial endowment to offset the difference. Participants decide how to explore a one‐dimensional space, receiving payoffs based on their location each period. We predict and find that participants are motivated to avoid losses, which increases exploration when they are incurring losses but decreases exploration when they face …

Consumer Psychology Review

Disclosure is a pervasive behavior critical to the human experience, and has been the focus of both empirical research and theorizing by social and consumer psychologists. This work has typically characterized disclosure as a deliberate and strategic act, the product of a careful weighing of costs and benefits. In this paper, we argue that, although some disclosure can be deliberate and “rational,” much of it exhibits drive‐like qualities. We review evidence to suggest that, much like other drive states (e.g., hunger), the desire to disclose can be visceral, driven by emotions and physical arousal, and satiated through the act of disclosing. And just as more basic drives evolved to motivate adaptive action but can engender maladaptive behavior (e.g., over‐eating), disclosures can likewise be impulsive and ultimately regrettable. We propose a dual‐process model that encompasses both viscerally driven and deliberate …

Available at SSRN 4358786

New technologies increase opportunities and pressures to learn new skills. Research on" desirable difficulties" has shown, however, that learners are often poor judges of the strategies for learning-eg, they fail to appreciate that spacing experiences with a task out over time will enhance learning. Focusing specifically on the choice of task difficulty, we examine whether learners choose tasks of appropriate difficulty to maximize their own learning, and if not (as we find), why they fail to do so. In Study 1 (N= 1,788) participants chose practice tasks that, on average, were easier than what would maximize learning. Studies 2 (N= 1,913) and 3 (N= 1,164), which examine whether this pattern of choice is best explained by preferences or beliefs, reveal that learners choose easier practice not because they are effort-averse, but because, on average, they hold incorrect beliefs about what level of task difficulty will maximize learning.

PLOS-ONE

People frequently face decisions that require making inferences about withheld information. The advent of large language models coupled with conversational technology, e.g., Alexa, Siri, Cortana, and the Google Assistant, is changing the mode in which people make these inferences. We demonstrate that conversational modes of information provision, relative to traditional digital media, result in more critical responses to withheld information, including: (1) a reduction in evaluations of a product or service for which information is withheld and (2) an increased likelihood of recalling that information was withheld. These effects are robust across multiple conversational modes: a recorded phone conversation, an unfolding chat conversation, and a conversation script. We provide further evidence that these effects hold for conversations with the Google Assistant, a prominent conversational technology. The experimental results point to participants’ intuitions about why the information was withheld as the driver of the effect.

PLOS-ONE

People frequently face decisions that require making inferences about withheld information. The advent of large language models coupled with conversational technology, e.g., Alexa, Siri, Cortana, and the Google Assistant, is changing the mode in which people make these inferences. We demonstrate that conversational modes of information provision, relative to traditional digital media, result in more critical responses to withheld information, including: (1) a reduction in evaluations of a product or service for which information is withheld and (2) an increased likelihood of recalling that information was withheld. These effects are robust across multiple conversational modes: a recorded phone conversation, an unfolding chat conversation, and a conversation script. We provide further evidence that these effects hold for conversations with the Google Assistant, a prominent conversational technology. The experimental results point to participants’ intuitions about why the information was withheld as the driver of the effect.

Plos-One

Objectives Existing literature on how employment loss affects depression has struggled to address potential endogeneity bias caused by reverse causality. The COVID-19 pandemic offers a unique natural experiment because the source of unemployment is very likely to be exogenous to the individual. This study assessed the effect of job loss and job furlough on the mental health of individuals in South Africa during the COVID-19 pandemic. Data and methods The data for the study came from the first and second waves of the national survey, the National Income Dynamics-Coronavirus Rapid Mobile Survey (NIDS-CRAM), conducted during May-June and July-August 2020, respectively. The sample for NIDS-CRAM was drawn from an earlier national survey, conducted in 2017, which had collected data on mental health. Questions on depressive symptoms during the lockdown were asked in Wave 2 of NIDS-CRAM, using a 2-question version of the Patient Health Questionnaire (PHQ-2). The PHQ-2 responses (0–6 on the discrete scale) were regrouped into four categories making the ordered logit regression model the most suited for assessing the impact of employment status on depressive symptoms. Results The study revealed that adults who retained paid employment during the COVID-19 lockdown had significantly lower depression scores than adults who lost employment. The benefits of employment also accumulated over time, underscoring the effect of unemployment duration on mental health. The analysis revealed no mental health benefits to being furloughed (on unpaid leave), but paid leave had a strong and significant positive …

Plos-One

Objectives Existing literature on how employment loss affects depression has struggled to address potential endogeneity bias caused by reverse causality. The COVID-19 pandemic offers a unique natural experiment because the source of unemployment is very likely to be exogenous to the individual. This study assessed the effect of job loss and job furlough on the mental health of individuals in South Africa during the COVID-19 pandemic. Data and methods The data for the study came from the first and second waves of the national survey, the National Income Dynamics-Coronavirus Rapid Mobile Survey (NIDS-CRAM), conducted during May-June and July-August 2020, respectively. The sample for NIDS-CRAM was drawn from an earlier national survey, conducted in 2017, which had collected data on mental health. Questions on depressive symptoms during the lockdown were asked in Wave 2 of NIDS-CRAM, using a 2-question version of the Patient Health Questionnaire (PHQ-2). The PHQ-2 responses (0–6 on the discrete scale) were regrouped into four categories making the ordered logit regression model the most suited for assessing the impact of employment status on depressive symptoms. Results The study revealed that adults who retained paid employment during the COVID-19 lockdown had significantly lower depression scores than adults who lost employment. The benefits of employment also accumulated over time, underscoring the effect of unemployment duration on mental health. The analysis revealed no mental health benefits to being furloughed (on unpaid leave), but paid leave had a strong and significant positive …

Plos-One

Objectives Existing literature on how employment loss affects depression has struggled to address potential endogeneity bias caused by reverse causality. The COVID-19 pandemic offers a unique natural experiment because the source of unemployment is very likely to be exogenous to the individual. This study assessed the effect of job loss and job furlough on the mental health of individuals in South Africa during the COVID-19 pandemic. Data and methods The data for the study came from the first and second waves of the national survey, the National Income Dynamics-Coronavirus Rapid Mobile Survey (NIDS-CRAM), conducted during May-June and July-August 2020, respectively. The sample for NIDS-CRAM was drawn from an earlier national survey, conducted in 2017, which had collected data on mental health. Questions on depressive symptoms during the lockdown were asked in Wave 2 of NIDS-CRAM, using a 2-question version of the Patient Health Questionnaire (PHQ-2). The PHQ-2 responses (0–6 on the discrete scale) were regrouped into four categories making the ordered logit regression model the most suited for assessing the impact of employment status on depressive symptoms. Results The study revealed that adults who retained paid employment during the COVID-19 lockdown had significantly lower depression scores than adults who lost employment. The benefits of employment also accumulated over time, underscoring the effect of unemployment duration on mental health. The analysis revealed no mental health benefits to being furloughed (on unpaid leave), but paid leave had a strong and significant positive …