The placebo effect: To explore or to exploit?

How people choose between options with differing outcomes (explore-exploit) is a central question to understanding human behaviour. However, the standard explore-exploit paradigm relies on gamified tasks with low-stake outcomes. Consequently, little is known about decision making for biologically-relevant stimuli. Here, we combined placebo and explore-exploit paradigms to examine detection and selection of the most effective treatment in a pain model. During conditioning, where 'optimal' and 'suboptimal' sham-treatments were paired with a reduction in electrical pain stimulation, participants learnt which treatment most successfully reduced pain. Modelling participant responses revealed three important findings. First, participants' choices reflected both directed and random exploration. Second, expectancy modulated pain - indicative of recursive placebo effects. Third, individual differences in terms of expectancy during conditioning predicted placebo effects during a subsequent test phase. These findings reveal directed and random exploration when the outcome is biologically-relevant. Moreover, this research shows how placebo and explore-exploit literatures can be unified.

How people learn about causal influence when there are many possible causes: A model based on informative transitions.

Four experiments tested how people learn cause-effect relations when there are many possible causes of an effect. When there are many cues, even if all the cues together strongly predict the effect, the bivariate relation between each individual cue and the effect can be weak, which can make it difficult to detect the influence of each cue. We hypothesized that when detecting the influence of a cue, in addition to learning from the states of the cues and effect (e.g., a cue is present and the effect is present), which is hypothesized by multiple existing theories of learning, participants would also learn from transitions - how the cues and effect change over time (e.g., a cue turns on and the effect turns on). We found that participants were better able to identify positive and negative cues in an environment in which only one cue changed from one trial to the next, compared to multiple cues changing (Experiments 1A, 1B). Within a single learning sequence, participants were also more likely to update their beliefs about causal strength when one cue changed at a time ('one-change transitions') than when multiple cues changed simultaneously (Experiment 2). Furthermore, learning was impaired when the trials were grouped by the state of the effect (Experiment 3) or when the trials were grouped by the state of a cue (Experiment 4), both of which reduce the number of one-change transitions. We developed a modification of the Rescorla-Wagner algorithm to model this 'Informative Transitions' learning processes.

Medication adherence as a learning process: insights from cognitive psychology.

Non-adherence to medications is one of the largest contributors to sub-optimal health outcomes. Many theories of adherence include a 'value-expectancy' component in which a patient decides to take a medication partly based on expectations about whether it is effective, necessary, and tolerable. We propose reconceptualising this common theme as a kind of 'causal learning' - the patient learns whether a medication is effective, necessary, and tolerable, from experience with the medication. We apply cognitive psychology theories of how people learn cause-effect relations to elaborate this causal-learning challenge. First, expectations and impressions about a medication and beliefs about how a medication works, such as delay of onset, can shape a patient's perceived experience with the medication. Second, beliefs about medications propagate both 'top-down' and 'bottom-up', from experiences with specific medications to general beliefs about medications and vice versa. Third, non-adherence can interfere with learning about a medication, because beliefs, adherence, and experience with a medication are connected in a cyclic learning problem. We propose that by conceptualising non-adherence as a causal-learning process, clinicians can more effectively address a patient's misconceptions and biases, helping the patient develop more accurate impressions of the medication.

Physician Bayesian updating from personal beliefs about the base rate and likelihood ratio.

Whether humans can accurately make decisions in line with Bayes' rule has been one of the most important yet contentious topics in cognitive psychology. Though a number of paradigms have been used for studying Bayesian updating, rarely have subjects been allowed to use their own preexisting beliefs about the prior and the likelihood. A study is reported in which physicians judged the posttest probability of a diagnosis for a patient vignette after receiving a test result, and the physicians' posttest judgments were compared to the normative posttest calculated from their own beliefs in the sensitivity and false positive rate of the test (likelihood ratio) and prior probability of the diagnosis. On the one hand, the posttest judgments were strongly related to the physicians' beliefs about both the prior probability as well as the likelihood ratio, and the priors were used considerably more strongly than in previous research. On the other hand, both the prior and the likelihoods were still not used quite as much as they should have been, and there was evidence of other nonnormative aspects to the updating, such as updating independent of the likelihood beliefs. By focusing on how physicians use their own prior beliefs for Bayesian updating, this study provides insight into how well experts perform probabilistic inference in settings in which they rely upon their own prior beliefs rather than experimenter-provided cues. It suggests that there is reason to be optimistic about experts' abilities, but that there is still considerable need for improvement.

Bayesian reasoning in residents' preliminary diagnoses.

Whether and when humans in general, and physicians in particular, use their beliefs about base rates in Bayesian reasoning tasks is a long-standing question. Unfortunately, previous research on whether doctors use their beliefs about the prevalence of diseases in diagnostic judgments has critical limitations. In this study, we assessed whether residents' beliefs about the prevalence of a disease are associated with their judgments of the likelihood of the disease in diagnosis, and whether residents' beliefs about the prevalence of diseases change across the 3 years of residency. Residents were presented with five ambiguous vignettes typical of patients presenting on the inpatient general medicine services. For each vignette, the residents judged the likelihood of five or six possible diagnoses. Afterward, they judged the prevalence within the general medicine services of all the diseases in the vignettes. Most importantly, residents who believed a disease to be more prevalent tended to rate the disease as more likely in the vignette cases, suggesting a rational tendency to incorporate their beliefs about disease prevalence into their diagnostic likelihood judgments. In addition, the residents' prevalence judgments for each disease were assessed over the 3 years of residency. The precision of the prevalence estimates increased across the 3 years of residency, though the accuracy of the prevalence estimates did not. These results imply that residents do have a rational tendency to use prevalence beliefs for diagnosis, and this finding also contributes to a larger question of whether humans intuitively use base rates for making judgments.

Reasoning about causal relationships: Inferences on causal networks.

Over the last decade, a normative framework for making causal inferences, Bayesian Probabilistic Causal Networks, has come to dominate psychological studies of inference based on causal relationships. The following causal networks-[X→Y→Z, X←Y→Z, X→Y←Z]-supply answers for questions like, "Suppose both X and Y occur, what is the probability Z occurs?" or "Suppose you intervene and make Y occur, what is the probability Z occurs?" In this review, we provide a tutorial for how normatively to calculate these inferences. Then, we systematically detail the results of behavioral studies comparing human qualitative and quantitative judgments to the normative calculations for many network structures and for several types of inferences on those networks. Overall, when the normative calculations imply that an inference should increase, judgments usually go up; when calculations imply a decrease, judgments usually go down. However, 2 systematic deviations appear. First, people's inferences violate the Markov assumption. For example, when inferring Z from the structure X→Y→Z, people think that X is relevant even when Y completely mediates the relationship between X and Z. Second, even when people's inferences are directionally consistent with the normative calculations, they are often not as sensitive to the parameters and the structure of the network as they should be. We conclude with a discussion of productive directions for future research.

Effect of grouping of evidence types on learning about interactions between observed and unobserved causes.

When a cause interacts with unobserved factors to produce an effect, the contingency between the observed cause and effect cannot be taken at face value to infer causality. Yet it would be computationally intractable to consider all possible unobserved, interacting factors. Nonetheless, 6 experiments found that people can learn about an unobserved cause participating in an interaction with an observed cause when the unobserved cause is stable over time. Participants observed periods in which a cause and effect were associated followed by periods of the opposite association ("grouped condition"). Rather than concluding a complete lack of causality, participants inferred that the observed cause does influence the effect (Experiment 1), and they gave higher causal strength estimates when there were longer periods during which the observed cause appeared to influence the effect (Experiment 2). Consistent with these results, when the trials were grouped, participants inferred that the observed cause interacted with an unobserved cause (Experiments 3 and 4). Indeed, participants could even make precise predictions about the pattern of interaction (Experiments 5 and 6). Implications for theories of causal reasoning are discussed.

Causal learning about tolerance and sensitization.

We introduce two abstract, causal schemata used during causal learning. (1) Tolerance is when an effect diminishes over time, as an entity is repeatedly exposed to the cause (e.g., a person becoming tolerant to caffeine). (2) Sensitization is when an effect intensifies over time, as an entity is repeatedly exposed to the cause (e.g., an antidepressant becoming more effective through repeated use). In Experiment 1, participants observed either of these cause-effect data patterns unfolding over time and exhibiting the tolerance or sensitization schemata. Participants inferred stronger causal efficacy and made more confident and more extreme predictions about novel cases than in a condition with the same data appearing in a random order over time. In Experiment 2, the same tolerance/sensitization scenarios occurred either within one entity or across many entities. In the many-entity conditions, when the schemata were violated, participants made much weaker inferences. Implications for causal learning are discussed.