Rational use of cognitive resources in human planning

Callaway, F., van Opheusden, B., Gul, S. et al. 

Nat Hum Behav 6, 1112–1125 (2022).

https://doi.org/10.1038/s41562-022-01332-8

Abstract

Making good decisions requires thinking ahead, but the huge number of actions and outcomes one could consider makes exhaustive planning infeasible for computationally constrained agents, such as humans. How people are nevertheless able to solve novel problems when their actions have long-reaching consequences is thus a long-standing question in cognitive science. To address this question, we propose a model of resource-constrained planning that allows us to derive optimal planning strategies. We find that previously proposed heuristics such as best-first search are near optimal under some circumstances but not others. In a mouse-tracking paradigm, we show that people adapt their planning strategies accordingly, planning in a manner that is broadly consistent with the optimal model but not with any single heuristic model. We also find systematic deviations from the optimal model that might result from additional cognitive constraints that are yet to be uncovered.

Discussion

In this paper, we proposed a rational model of resource-constrained planning and compared the predictions of the model to human behaviour in a process-tracing paradigm. Our results suggest that human planning strategies are highly adaptive in ways that previous models cannot capture. In Experiment 1, we found that the optimal planning strategy in a generic environment resembled best-first search with a relative stopping rule. Participant behaviour was also consistent with such a strategy. However, the optimal planning strategy depends on the structure of the environment. Thus, in Experiments 2 and 3, we constructed six environments in which the optimal strategy resembled different classical search algorithms (best-first, breadth-first, depth-first and backward search). In each case, participant behaviour matched the environment-appropriate algorithm, as the optimal model predicted.

The idea that people use heuristics that are jointly adapted to environmental structure and computational limitations is not new. First popularized by Herbert Simon, it has more recently been championed in ecological rationality, which generally takes the approach of identifying computationally frugal heuristics that make accurate choices in certain environments. However, while ecological rationality explicitly rejects the notion of optimality, our approach embraces it, identifying heuristics that maximize an objective function that includes both external utility and internal cognitive cost. Supporting our approach, we found that the optimal model explained human planning behaviour better than flexible combinations of previously proposed planning heuristics in seven of the eight environments we considered (Supplementary Table 1).

The info is here.

Filling the gaps: Cognitive control as a critical lens for understanding mechanisms of value-based decision-making.

Frömer, R., & Shenhav, A. (2021, May 17). 

https://doi.org/10.31234/osf.io/dnvrj

Abstract

While often seeming to investigate rather different problems, research into value-based decision making and cognitive control have historically offered parallel insights into how people select thoughts and actions. While the former studies how people weigh costs and benefits to make a decision, the latter studies how they adjust information processing to achieve their goals. Recent work has highlighted ways in which decision-making research can inform our understanding of cognitive control. Here, we provide the complementary perspective: how cognitive control research has informed understanding of decision-making. We highlight three particular areas of research where this critical interchange has occurred: (1) how different types of goals shape the evaluation of choice options, (2) how people use control to adjust how they make their decisions, and (3) how people monitor decisions to inform adjustments to control at multiple levels and timescales. We show how adopting this alternate viewpoint offers new insight into the determinants of both decisions and control; provides alternative interpretations for common neuroeconomic findings; and generates fruitful directions for future research.

Highlights

•  We review how taking a cognitive control perspective provides novel insights into the mechanisms of value based choice.

•  We highlight three areas of research where this critical interchange has occurred:

      (1) how different types of goals shape the evaluation of choice options,

      (2) how people use control to adjust how they make their decisions, and

      (3) how people monitor decisions to inform adjustments to control at multiple levels and timescales.

From Exerting Control Beyond Our Current Choice

We have so far discussed choices the way they are typically studied:in isolation. However, we don’t make choices in a vacuum, and our current choices depend on previous choices we have made (Erev & Roth, 2014; Keung, Hagen, & Wilson, 2019; Talluri et al., 2020; 618Urai, Braun, & Donner, 2017; Urai, de Gee, Tsetsos, & Donner, 2019). One natural way in which choices influence each other is through learning about the options, where the evaluations of the outcome of one choice refines the expected value (incorporating range and probability) assigned to that option in future choices (Fontanesi, Gluth, et al., 2019; Fontanesi, Palminteri, et al., 2019; Miletic et al., 2021).  Here we focus on a different, complementary way, central to cognitive control research, where evaluations of the process of ongoing and past choices inform the process of future choices(Botvinick et al., 1999; Bugg, Jacoby, & Chanani, 2011; Verguts, Vassena, & Silvetti, 2015). In cognitive control research, these choice evaluations and their influence on subsequent adaptation are studied under the umbrella of performance monitoring (Carter et al., 1998; Ullsperger, Fischer, Nigbur, & Endrass, 2014). Unlike option-based learning, performance monitoring influences not only which options are chosen, but also how subsequent choices are made. It also informs higher order decisions about strategy and task selection(Fig. 6305A).

The research is here.

AI Is Not Similar To Human Intelligence. Thinking So Could Be Dangerous

Elizabeth Fernandez
forbes.com
Originally posted 30 Nov 19

Here is an excerpt:

No doubt, these algorithms are powerful, but to think that they “think” and “learn” in the same way as humans would be incorrect, Watson says. There are many differences, and he outlines three.

The first - DNNs are easy to fool. For example, imagine you have a picture of a banana. A neural network successfully classifies it as a banana. But it’s possible to create a generative adversarial network that can fool your DNN. By adding a slight amount of noise or another image besides the banana, your DNN might now think the picture of a banana is a toaster. A human could not be fooled by such a trick. Some argue that this is because DNNs can see things humans can’t, but Watson says, “This disconnect between biological and artificial neural networks suggests that the latter lack some crucial component essential to navigating the real world.”

Secondly, DNNs need an enormous amount of data to learn. An image classification DNN might need to “see” thousands of pictures of zebras to identify a zebra in an image. Give the same test to a toddler, and chances are s/he could identify a zebra, even one that’s partially obscured, by only seeing a picture of a zebra a few times. Humans are great “one-shot learners,” says Watson. Teaching a neural network, on the other hand, might be very difficult, especially in instances where data is hard to come by.

Thirdly, neural nets are “myopic”. They can see the trees, so to speak, but not the forest. For example, a DNN could successfully label a picture of Kim Kardashian as a woman, an entertainer, and a starlet. However, switching the position of her mouth and one of her eyes actually improved the confidence of the DNN’s prediction. The DNN didn’t see anything wrong with that image. Obviously, something is wrong here. Another example - a human can say “that cloud looks like a dog”, whereas a DNN would say that the cloud is a dog.

The info is here.

Exploiting Risk–Reward Structures in Decision Making under Uncertainty

Christina Leuker Thorsten Pachur Ralph Hertwig Timothy Pleskac
PsyArXiv Preprints
Posted December 21, 2017

Abstract

People often have to make decisions under uncertainty — that is, in situations where the probabilities of obtaining a reward are unknown or at least difficult to ascertain. Because outside the laboratory payoffs and probabilities are often correlated, one solution to this problem might be to infer the probability from the magnitude of the potential reward. Here, we investigated how the mind may implement such a solution: (1) Do people learn about risk–reward relationships from the environment—and if so, how? (2) How do learned risk–reward relationships impact preferences in decision-making under uncertainty? Across three studies (N = 352), we found that participants learned risk–reward relationships after being exposed to choice environments with a negative, positive, or uncorrelated risk–reward relationship. They learned the associations both from gambles with explicitly stated payoffs and probabilities (Experiments 1 & 2) and from gambles about epistemic
events (Experiment 3). In subsequent decisions under uncertainty, participants exploited the learned association by inferring probabilities from the magnitudes of the payoffs. This inference systematically influenced their preferences under uncertainty: Participants who learned a negative risk–reward relationship preferred the uncertain option over a smaller sure option for low payoffs, but not for high payoffs. This pattern reversed in the positive condition and disappeared in the uncorrelated condition. This adaptive change in preferences is consistent with the use of the risk–reward heuristic.

From the Discussion Section:

Risks and rewards are the pillars of preference. This makes decision making under uncertainty a vexing problem as one of those pillars—the risks, or probabilities—is missing (Knight, 1921; Luce & Raiffa, 1957). People are commonly thought to deal with this problem by intuiting subjective probabilities from their knowledge and memory (Fox & Tversky, 1998; Tversky & Fox, 1995) or by estimating statistical probabilities from samples of information (Hertwig & Erev, 2009). Our results support another ecologically grounded solution, namely, that people estimate the missing probabilities from their immediate choice environments via their learned risk–reward relationships.

The research is here.

Are morally good actions ever free?

Cory J. Clark, Adam Shniderman, Jamie Luguri, Roy Baumeister, and Peter Ditto
SSRN Electronic Journal, August 2017

Abstract

A large body of work has demonstrated that people ascribe more responsibility to morally bad actions than both morally good and morally neutral ones, creating the impression that people do not attribute responsibility to morally good actions. The present work demonstrates that this is not so: People attributed more free will to morally good actions than morally neutral ones (Studies 1a-1b). Studies 2a-2b distinguished the underlying motives for ascribing responsibility to morally good and bad actions. Free will ascriptions for morally bad actions were driven predominantly by affective punitive responses. Free will judgments for morally good actions were similarly driven by affective reward responses, but also less affectively-charged and more pragmatic considerations (the perceived utility of reward, normativity of the action, and willpower required to perform the action). Responsibility ascriptions to morally good actions may be more carefully considered, leading to generally weaker, but more contextually-sensitive free will judgments.

The research is here.

Approach-Induced Biases in Human Information Sampling

Laurence T. Hunt and others
PLOS Biology
Published: November 10, 2016

Abstract

Information sampling is often biased towards seeking evidence that confirms one’s prior beliefs. Despite such biases being a pervasive feature of human behavior, their underlying causes remain unclear. Many accounts of these biases appeal to limitations of human hypothesis testing and cognition, de facto evoking notions of bounded rationality, but neglect more basic aspects of behavioral control. Here, we investigated a potential role for Pavlovian approach in biasing which information humans will choose to sample. We collected a large novel dataset from 32,445 human subjects, making over 3 million decisions, who played a gambling task designed to measure the latent causes and extent of information-sampling biases. We identified three novel approach-related biases, formalized by comparing subject behavior to a dynamic programming model of optimal information gathering. These biases reflected the amount of information sampled (“positive evidence approach”), the selection of which information to sample (“sampling the favorite”), and the interaction between information sampling and subsequent choices (“rejecting unsampled options”). The prevalence of all three biases was related to a Pavlovian approach-avoid parameter quantified within an entirely independent economic decision task. Our large dataset also revealed that individual differences in the amount of information gathered are a stable trait across multiple gameplays and can be related to demographic measures, including age and educational attainment. As well as revealing limitations in cognitive processing, our findings suggest information sampling biases reflect the expression of primitive, yet potentially ecologically adaptive, behavioral repertoires. One such behavior is sampling from options that will eventually be chosen, even when other sources of information are more pertinent for guiding future action.

The article is here.

From Good Institutions to Good Norms: Top-Down Incentives to Cooperate Foster Prosociality But Not Norm Enforcement

Michael N Stagnaro, Antonio A. Arechar, & David G. Rand
Social Science Research Network

Abstract:  

What makes people willing to pay costs to help others, and to punish others’ selfishness? And why does the extent of such behaviors vary markedly across cultures? To shed light on these questions, we explore the role of formal institutions in shaping individuals’ prosociality and punishment. In Study 1 (N=707), we found that the quality of the institutions that participants were exposed to in daily life was positively associated with giving in a Dictator Game, but had little relationship with punishment in a Third-Party Punishment Game. In Study 2 (N=516), we investigated causality by experimentally manipulating institutional quality using a centralized punishment institution applied to a repeated Public Goods Game. Consistent with Study 1’s correlational results, we found that high institutional quality led to significantly more prosociality in a subsequent Dictator Game, but did not have a significant overall effect on subsequent punishment. Thus we present convergent evidence that the quality of institutions one is exposed to “spills over” to affect subsequent prosociality, but not punishment. These findings support a theory of social heuristics, suggest boundary conditions on spillover effects of cooperation, and demonstrate the power of effective institutions for instilling habits of virtue and creating cultures of cooperation.

The article is here.

On the misguided pursuit of happiness and ethical decision making: The roles of focalism and the impact bias in unethical and selfish behavior

Laura J. Noval
Organizational Behavior and Human Decision Processes
Volume 133, March 2016, Pages 1–16

Abstract

An important body of research in the field of behavioral ethics argues that individuals behave unethically and selfishly because they want to obtain desired outcomes, such as career advancement and monetary rewards. Concurrently, a large body of literature in social psychology has shown that the subjective value of an outcome is determined by its anticipated emotional impact. Such impact has been consistently found to be overestimated both in its intensity and in its duration (i.e. impact bias) due to focalism (i.e. excessive focus on the desired outcome). Across four empirical studies, this investigation demonstrates that reducing focalism and thereby attenuating the impact bias in regards to desired outcomes decreases people’s tendency to engage in both unethical and selfish behavior to obtain those outcomes.

Highlights

• Individuals engage in unethical and selfish behavior to obtain desired outcomes, such as monetary or career rewards.
• The anticipated emotional impact of the outcomes individuals seek to obtain is overestimated (i.e. impact bias).
• The impact bias results from focalism (i.e. excessive focus on an outcome).
• In four studies, focalism and the impact bias about desired outcomes were experimentally reduced.
• The focalism reduction resulted in a decreased tendency of individuals to engage in unethical and selfish behavior.

The article is here.

People Don't Actually Want Equality

By Paul Bloom
The Atlantic
Originally published on October 22, 2015

Here is an excerpt:

Can Frankfurt really be right that people don’t value economic equality for its own sake? Many scholars believe otherwise. The primatologist Frans de Waal sums up a popular view when he writes: “Robin Hood had it right. Humanity’s deepest wish is to spread the wealth.”

In support of de Waal, researchers have found that if you ask children to distribute items to strangers, they are strongly biased towards equal divisions, even in extreme situations. The psychologists Alex Shaw and Kristina Olson told children between the ages of six and eight about two boys, Dan and Mark, who had cleaned up their room and were to be rewarded with erasers—but there were five of them, so an even split was impossible. Children overwhelmingly reported that the experimenter should throw away the fifth eraser rather than establish an unequal division. They did so even if they could have given the eraser to Dan or Mark without the other one knowing, so they couldn’t have been worrying about eliciting anger or envy.

It might seem as though these responses reflect a burning desire for equality, but more likely they reflect a wish for fairness. It is only because Dan and Mark did the same work that they should get the same reward. And so when Shaw and Olson told the children “Dan did more work than Mark,” they were quite comfortable giving three to Dan and two to Mark. In other words, they were fine with inequality, so long as it was fair.

The entire article is here.

Ethical Leadership: Meta-Analytic Evidence of Criterion-Related and Incremental Validity

By Thomas W. H. Ng and Daniel C. Feldman
J Appl Psychol. 2014 Nov 24

Abstract

This study examines the criterion-related and incremental validity of ethical leadership (EL) with meta-analytic data. Across 101 samples published over the last 15 years (N = 29,620), we observed that EL demonstrated acceptable criterion-related validity with variables that tap followers' job attitudes, job performance, and evaluations of their leaders. Further, followers' trust in the leader mediated the relationships of EL with job attitudes and performance. In terms of incremental validity, we found that EL significantly, albeit weakly in some cases, predicted task performance, citizenship behavior, and counterproductive work behavior-even after controlling for the effects of such variables as transformational leadership, use of contingent rewards, management by exception, interactional fairness, and destructive leadership. The article concludes with a discussion of ways to strengthen the incremental validity of EL.

The entire article is here.

Finding Risks, Not Answers, in Gene Tests

By Denise Grady and Andrew Pollack
The New York Times
Originally published September 22, 2014

Jennifer was 39 and perfectly healthy, but her grandmother had died young from breast cancer, so she decided to be tested for mutations in two genes known to increase risk for the disease.

When a genetic counselor offered additional tests for 20 other genes linked to various cancers, Jennifer said yes. The more information, the better, she thought.

The results, she said, were “surreal.” She did not have mutations in the breast cancer genes, but did have one linked to a high risk of stomach cancer. In people with a family history of the disease, that mutation is considered so risky that patients who are not even sick are often advised to have their stomachs removed. But no one knows what the finding might mean in someone like Jennifer, whose family has not had the disease.

The entire article is here.

Intuitive Prosociality

By Jamil Zaki and Jason P. Mitchell
Current Directions in Psychological Science 22(6) 466–470
DOI: 10.1177/0963721413492764

Abstract

Prosocial behavior is a central feature of human life and a major focus of research across the natural and social sciences. Most theoretical models of prosociality share a common assumption: Humans are instinctively selfish, and prosocial behavior requires exerting reflective control over these basic instincts. However, findings from several scientific disciplines have recently contradicted this view. Rather than requiring control over instinctive selfishness, prosocial behavior appears to stem from processes that are intuitive, reflexive, and even automatic. These observations suggest that our understanding of prosociality should be revised to include the possibility that, in many cases, prosocial behavior—instead of requiring active control over our impulses—represents an impulse of its own.

Click here for accessing the article, behind a paywall.