Psychology and AI at a Crossroads: How Might Complex Systems Explain Themselves?

A challenge in building useful artificial intelligence (AI) systems is that people need to understand how they work in order to achieve appropriate trust and reliance. This has become a topic of considerable interest, manifested as a surge of research on Explainable AI (XAI). Much of the research assumes a model in which the AI automatically generates an explanation and presents it to the user, whose understanding of the explanation leads to better performance. Psychological research on explanatory reasoning shows that this is a limited model. The design of XAI systems must be fully informed by a model of cognition and a model of pedagogy, based on empirical evidence of what happens when people try to explain complex systems to other people and what happens as people try to reason out how a complex system works. In this article we discuss how and why C. S. Peirce's notion of abduction is a best model for XAI. Peirce's notion of abduction as an exploratory activity can be regarded as supported by virtue of its concordance with models of expert reasoning that have been developed by modern applied cognitive psychologists.

Tribute to E. J. Capaldi: Celebration of a Psychological Scientist

E. J. (John) Capaldi (1928–2020) made numerous contributions to experimental psychology in his long career at the University of Texas at Austin and Purdue University. He was a pioneer in the area of animal learning and cognition, known for his sequential theory of partial reinforcement extinction effects. His research in this area and in memory and counting phenomena was conducted for the most part with rats running in straight alley mazes under various sequences of trial outcomes (e.g., reward, nonreward, variations in reward size). John's other interests included the sequential theory applied to Pavlovian conditioning, evolutionary theory, and psychology of science. More generally, John was a scientist to the core who served as a role model to those who knew him. His career centered on the values of science and rigorous critical thinking based in empirical data, coupled with a curiosity and openness to different and new ideas in psychology.

C. S. Peirce’s Forgotten but Enduring Relevance to Psychological Science

Charles Sanders Peirce (1839-1914) was one of the most polymathically brilliant scientific thinkers in American history. He was also arguably the first American experimental psychologist, strongly influencing some of the nation’s earliest modern psychology pioneers. Yet partly because of the lasting effects of personal scandals and powerful enemies, he has been almost entirely forgotten by the broader field of psychology. This article aims to briefly reintroduce Peirce as a historically important figure in psychology to a general audience and highlight a few ways in which his trailblazing perspectives point to pervasive deficiencies and opportunities in psychological science. First, his pioneering writings on the economy of research call on us to consider how diminishing returns may plague our research programs, potentially leading to waste of time, money, and intellectual labor in our communities of inquiry. Second, his pragmatic maxim of clarity for intellectual constructs provides a compelling framework in which to understand how jingle and jangle fallacies often undercut the cumulativeness of our field. We contend that Peirce’s work, much of it still unpublished, is a rich resource for psychological scientists across many domains.

Challenges Facing Evidence-Based Prevention: Incorporating an Abductive Theory of Method

Current systems used to determine whether prevention programs are "evidence-based" rely on the logic of deductive reasoning. This reliance has fostered implementation of strategies with explicitly stated evaluation criteria used to gauge program validity and suitability for dissemination. Frequently, investigators resort to the randomized controlled trial (RCT) combined with null hypothesis significance testing (NHST) as a means to rule out competing hypotheses and determine whether an intervention works. The RCT design has achieved success across numerous disciplines but is not without limitations. We outline several issues that question allegiance to the RCT, NHST, and the hypothetico-deductive method of scientific inquiry. We also discuss three challenges to the status of program evaluation including reproducibility, generalizability, and credibility of findings. As an alternative, we posit that extending current program evaluation criteria with principles drawn from an abductive theory of method (ATOM) can strengthen our ability to address these challenges and advance studies of drug prevention. Abductive reasoning involves working from observed phenomena to the generation of alternative explanations for the phenomena and comparing the alternatives to select the best possible explanation. We conclude that an ATOM can help increase the influence and impact of evidence-based prevention for population benefit.

Dialogically based approaches to “with” and “about” the other: Thoughts on Carl Rogers’ dilemma

Over 50 years ago, Carl Rogers wrote of his personal struggles with two conflicting self-voices, one striving to be empathically “with” another person, and the other seeking abstract, theoretical insights “about” the person. In more recent times, these two voices have come to represent seemingly entrenched, opposing camps within the domain of interpretive inquiry, an impasse which some suggest can only be resolved within a dialogical framework. However, there are important differences between dialogical perspectives in their potential for serving as such a framework. In this article, I take up the question of why a relational dialogism is required to organize and coordinate “with and about the other” and associated dualities. During the course of this discussion, I address meanings that reinforce the division and how these might be re-envisioned and then offer examples of interpretive research that is based in a relational dialogism with an emphasis on theoretical innovation.

A disconfirming strategy is not necessarily better than a confirming strategy

College students were presented with 4 events to be explained. Each event was accompanied by either a target explanation or a target and an alternative, and all explanations were causal mechanistic. For each event, participants were also presented with 12 pieces of information. Four of the 12 pieces were causally consistent with the target, 4 with the alternative, and 4 were neutral. Participants were assigned to one of 3 strategy conditions, depending on what they were explicitly told to do: confirm, disconfirm, or evaluate the target explanation. People told to disconfirm were more likely than people in the other conditions to cite, as strategy appropriate, information that was in fact neutral. In addition, across strategy conditions, the presence of an alternative decreased the likelihood of identifying, as appropriate to the strategy, information that was strategy inappropriate. Results are discussed in terms of inference to the best explanation and in terms of confirmation bias.