Decide
The value of mere completion
Benjamin Converse, Shelly Tsang & Marie Hennecke
Journal of Experimental Psychology: General, forthcoming
Abstract:
The positivity of goal completion is reinforced through everyday experiences of social praise and instrumental reward. Here we investigated whether, in line with this self-regulatory emphasis, people value completion opportunities in and of themselves. Across six experiments we found that adding an arbitrary completion opportunity to a lower-reward task increased the likelihood that participants would choose to work on that task over a higher-reward alternative that did not offer a completion opportunity. This occurred for extrinsic reward tradeoffs (Experiments 1, 3, 4, and 5) and intrinsic reward tradeoffs (Experiments 2 and 6), and it persisted even when participants explicitly noted the rewards of each task (Experiment 3). We sought but did not find evidence that the tendency is moderated by participants’ stable or momentary level of concern with monitoring multiple responsibilities (Experiments 4 and 5, respectively). We did find that the opportunity to complete the final step in a sequence was particularly attractive: Setting the lower-reward task closer to completion (but with completion still out of reach) did increase its choice share, but setting the lower-reward task with completion distinctly in reach increased its choice share even more (Experiment 6). Together, the experiments imply that people sometimes behave as if they value completion itself. In everyday life, the allure of mere completion may influence the tradeoffs people make when prioritizing their goals.
Combining Human Expertise with Artificial Intelligence: Experimental Evidence from Radiology
Nikhil Agarwal et al.
NBER Working Paper, July 2023
Abstract:
While Artificial Intelligence (AI) algorithms have achieved performance levels comparable to human experts on various predictive tasks, human experts can still access valuable contextual information not yet incorporated into AI predictions. Humans assisted by AI predictions could outperform both human-alone or AI-alone. We conduct an experiment with professional radiologists that varies the availability of AI assistance and contextual information to study the effectiveness of human-AI collaboration and to investigate how to optimize it. Our findings reveal that (i) providing AI predictions does not uniformly increase diagnostic quality, and (ii) providing contextual information does increase quality. Radiologists do not fully capitalize on the potential gains from AI assistance because of large deviations from the benchmark Bayesian model with correct belief updating. The observed errors in belief updating can be explained by radiologists’ partially underweighting the AI’s information relative to their own and not accounting for the correlation between their own information and AI predictions. In light of these biases, we design a collaborative system between radiologists and AI. Our results demonstrate that, unless the documented mistakes can be corrected, the optimal solution involves assigning cases either to humans or to AI, but rarely to a human assisted by AI.
When knowledge hurts: Humans are willing to receive pain for obtaining non-instrumental information
Stefan Bode et al.
Proceedings of the Royal Society: Biological Sciences, 12 July 2023
Abstract:
Humans and other animals value information that reduces uncertainty or leads to pleasurable anticipation, even if it cannot be used to gain tangible rewards or change outcomes. In exchange, they are willing to incur significant costs, sacrifice rewards or invest effort. We investigated whether human participants were also willing to endure pain -- a highly salient and aversive cost -- to obtain such information. Forty participants performed a computer-based task. On each trial, they observed a coin flip, with each side associated with different monetary rewards of varying magnitude. Participants could choose to endure a painful stimulus (low, moderate or high pain) to learn the outcome of the coin flip immediately. Importantly, regardless of their choice, winnings were always earned, rendering this information non-instrumental. Results showed that agents were willing to endure pain in exchange for information, with a lower likelihood of doing so as pain levels increased. Both higher average rewards and a larger variance between the two possible rewards independently increased the willingness to accept pain. Our results show that the intrinsic value of escaping uncertainty through non-instrumental information is sufficient to offset pain experiences, suggesting a shared mechanism through which these can be directly compared.
Art-ificial Intelligence: The Effect of AI Disclosure on Evaluations of Creative Content
Manav Raj, Justin Berg & Robert Seamans
Stanford Working Paper, March 2023
Abstract:
The emergence of generative AI technologies, such as OpenAI’s ChatGPT chatbot, has expanded the scope of tasks that AI tools can accomplish and enabled AI-generated creative content. In this study, we explore how disclosure regarding the use of AI in the creation of creative content affects human evaluation of such content. In a series of pre-registered experimental studies, we show that AI disclosure has no meaningful effect on evaluation either for creative or descriptive short stories, but that AI disclosure has a negative effect on evaluations for emotionally evocative poems written in the first person. We interpret this result to suggest that reactions to AI-generated content may be negative when the content is viewed as distinctly “human.” We discuss the implications of this work and outline planned pathways of research to better understand whether and when AI disclosure may affect the evaluation of creative content.
Is there a g in gunslinger? Cognitive predictors of firearms proficiency
Jeffrey Cucina et al.
Intelligence, July-August 2023
Abstract:
This study addressed a gap in the research literature by evaluating the validity of general mental ability (g) and personality test scores for prediction of firearms proficiency via shooting range performance, an entirely objective task-based criterion. It was hypothesized that mental ability test scores would be positively related to firearms proficiency based on past research in related areas (e.g., g predicts skill acquisition and training performance) and conceptual similarities between firearms proficiency and cognitive tasks. Using 4 datasets with a combined sample size of 22,525 individuals, this hypothesis was confirmed: g had operational validities ranging from .162 to .188 and logical reasoning had operational validities ranging from .179 to .268 after correcting for range restriction and criterion unreliability. Mental ability test scores predicted an entirely psychomotor criterion task: use of firearms to hit targets at a pre-determined level of accuracy. Most of the validity appears to be attributable to g, but a post hoc analysis indicated that writing ability acted as a suppressor (i.e., the validity of g increased when writing ability was included in a regression model). Conscientiousness was hypothesized to have a positive relationship with firearms performance and emotional stability was hypothesized to have positive linear and quadratic relationships. In contrast, it was observed that conscientiousness had a negative operational validity (−.079) and emotional stability lacked validity relative to the firearms proficiency criterion. The implications for individual differences research and practice are discussed.
The “curse of knowledge” when predicting others’ knowledge
Jonathan Tullis & Brennen Feder
Memory & Cognition, July 2023, Pages 1214–1234
Abstract:
To succeed in a social world, we must be able to accurately estimate what others know. For example, teachers must anticipate student knowledge to plan lessons and communicate effectively. Yet one’s own knowledge consistently contaminates estimates about others’ knowledge. We examine how one’s knowledge influences the calibration and resolution of participants’ estimates of novices’ knowledge. Across four experiments, participants studied trivia questions and estimated the percentage of novice participants who would know the answer across multiple study/estimation rounds. When participants were required to answer the question before estimating what novices would know, studying the facts impaired both the calibration and resolution of the estimates. Studying the facts reduced the validity of one’s experiences for predicting novices’ knowledge, and estimators utilized their own experiences less when predicting novices’ knowledge as they studied. Experimentally reducing reliance on one’s own knowledge did not improve the accuracy of estimates. The results suggest that learning impairs the accuracy of judgments of others’ knowledge, not because estimators rely too heavily on their own experiences, but because estimators lack diagnostic cues about others’ knowledge.
Social interoception: Perceiving events during cardiac afferent activity makes people more suggestible to other people's influence
Mariana von Mohr et al.
Cognition, forthcoming
Abstract:
Our judgements are often influenced by other people's views and opinions. Interoception also influences decision making, but little is known about its role in social influence and particularly, the extent to which other people may influence our decisions. Across two experiments, using different forms of social influence, participants judged the trustworthiness of faces presented either during the systolic phase of the cardiac cycle, when baroreceptors convey information from the heart to the brain, or during diastolic phase, when baroreceptors are quiescent. We quantified the extent to which participants changed their minds (as an index of social influence) following the social feedback, in order to compare two competing hypotheses. According to the Arousal-Confidence Hypothesis, cardiac signals create a context of heightened bodily arousal that increases confidence in perceptual judgements. People should, therefore, be less subject to social influence during systole. By contrast, according to the Uncertainty-Conformity Hypothesis, cardiac signals increase neural noise and sensory attenuation, such that people should display greater effects of social influence during systole, as they then underweight private interoceptive signals in favour of the external social information. Across two studies that used different kind of social interactions, we found that participants changed their minds more when faces were presented at systole. Our results, therefore, support the Uncertainly-Conformity hypothesis and highlight how cardiac afferent signals contribute to shape our social decision-making in different types of social interactions.
Memory for artwork is predictable
Trent Davis & Wilma Bainbridge
Proceedings of the National Academy of Sciences, 11 July 2023
Abstract:
Viewing art is often seen as a highly personal and subjective experience. However, are there universal factors that make a work of art memorable? We conducted three experiments, where we recorded online memory performance for 4,021 paintings from the Art Institute of Chicago, tested in-person memory after an unconstrained visit to the Art Institute, and obtained abstract attribute measures such as beauty and emotional valence for these pieces. Participants showed significant agreement in their memories both online and in-person, suggesting that pieces have an intrinsic “memorability” based solely on their visual properties that is predictive of memory in a naturalistic museum setting. Importantly, ResMem, a deep learning neural network designed to estimate image memorability, could significantly predict memory both online and in-person based on the images alone, and these predictions could not be explained by other low- or high-level attributes like color, content type, aesthetics, and emotion. A regression comprising ResMem and other stimulus factors could predict as much as half of the variance of in-person memory performance. Further, ResMem could predict the fame of a piece, despite having no cultural or historical knowledge. These results suggest that perceptual features of a painting play a major role in influencing its success, both in memory for a museum visit and in cultural memory over generations.
Re-Examining Moral Hazard Under Inattention: New Evidence from Behavioral Data in Auto Insurance
Yizhou Jin
University of Toronto Working Paper, May 2023
Abstract:
This paper uses novel sensor data to study drivers’ risky phone use behavior. The results challenge the conventional wisdom of moral hazard in insurance. We first identify handheld phone use behavior (“HPU”) and quantify its causal impact on accident likelihood (“riskiness”) using exhaustive fixed-effect models. We then find HPU to be risky but insensitive to both insurance coverage changes and weather shocks that increase its riskiness. This contradicts the prevailing theoretical prediction and empirical studies that have thus far relied on claims data alone. On the other hand, an experiment with a one-time text-message warning led to a persistent 15% HPU reduction. Drivers’ inattention to risk thus limits moral hazard.
Having multiple selves helps learning agents explore and adapt in complex changing worlds
Zack Dulberg et al.
Proceedings of the National Academy of Sciences, 11 July 2023
Abstract:
Satisfying a variety of conflicting needs in a changing environment is a fundamental challenge for any adaptive agent. Here, we show that designing an agent in a modular fashion as a collection of subagents, each dedicated to a separate need, powerfully enhanced the agent’s capacity to satisfy its overall needs. We used the formalism of deep reinforcement learning to investigate a biologically relevant multiobjective task: continually maintaining homeostasis of a set of physiologic variables. We then conducted simulations in a variety of environments and compared how modular agents performed relative to standard monolithic agents (i.e., agents that aimed to satisfy all needs in an integrated manner using a single aggregate measure of success). Simulations revealed that modular agents a) exhibited a form of exploration that was intrinsic and emergent rather than extrinsically imposed; b) were robust to changes in nonstationary environments, and c) scaled gracefully in their ability to maintain homeostasis as the number of conflicting objectives increased. Supporting analysis suggested that the robustness to changing environments and increasing numbers of needs were due to intrinsic exploration and efficiency of representation afforded by the modular architecture. These results suggest that the normative principles by which agents have adapted to complex changing environments may also explain why humans have long been described as consisting of “multiple selves.”