What do our sampling assumptions affect: how we encode data or how we reason from it?

Abstract

In describing how people generalize from observed samples of data to novel cases, theories of inductive inference have emphasized the learner's reliance on the contents of the sample. More recently, a growing body of literature suggests that different assumptions about how a data sample was generated can lead the learner to draw qualitatively different inferences on the basis of the same evidence. Yet relatively little is known about how and when these two sources of evidence are combined. For instance, do sampling assumptions affect how the sample contents are encoded, or is any influence exerted only at the point of retrieval when a decision is to be made? We report two experiments aimed at exploring this issue. By systematically varying both the sampling cover story and whether it is given before or after the training stimuli we are able to determine whether encoding or retrieval issues drive the impact of sampling assumptions. Across two experiments we find that the sampling cover story affects generalization when it is presented before the training stimuli, but not after, which we interpret in favor of an encoding account.

Does explaining the origins of misinformation improve the effectiveness of a given correction?

“…Instilling a sense of distrust about the misinformation source's motives may also be more effective at encoding than at retrieval (i.e., when misinformation is encoded rather than when the correction is presented). This explanation also fits with recent work showing that people are better able to incorporate information about constraints on a sample of evidence when it presented at encoding than when presented at retrieval (Ransom et al, 2021).…”

Who is sensitive to selection biases in inductive reasoning?

“…The latter condition should mean that the frames instructions were easier to retrieve for the generalization test. Ransom et al (in press), however found a weaker sampling frames effect in the ‘after’ compared with the ‘before’ condition. Hence, it seems unlikely that retrieval of frames instructions at the time of testing is a key factor in determining generalization performance.…”

“…The Ransom et al (in press) results imply that applying frames instructions as sample instances are encoded is crucial. Superior working memory ability memory may impact this process by making it easier to encode each sample instance and to use this information to update the probabilities of rival hypotheses about property extension (cf.…”

“…While we cannot discount this possibility, the results of another recent study speak against it. Ransom et al (in press) compared the effects on inductive inference of presenting category or property frames before the sample data (as in the current study) or after the sample data were observed (but before the generalization test). The latter condition should mean that the frames instructions were easier to retrieve for the generalization test.…”

Inferring the truth from deception: What can people learn from helpful and unhelpful information providers?

“…Much of the research investigating sampling assumptions has focused on how people reason from weak or strong samples. Under weak sampling, examples have been randomly sampled from the world and then labelled, while under strong sampling, they are randomly sampled from the category being learned (Anderson, 1981;Hendrickson et al, 2019;Navarro et al, 2012;Ransom et al, 2021;Tenenbaum & Griffiths, 2001). While this research is useful for beginning to think about what assumptions underlie people's ability to make rich inferences from sparse data, realistic scenarios encompass a range of assumptions that go beyond strong and weak alone.…”