Strategies of Rule Discovery in an Inference Task

Abstract

It has long been known that subjects in certain inference tasks will seek evidence which can confirm their present hypotheses, even in situations where disconfirmatory evidence could be more informative. We sought to alter this tendency in a series of experiments which employed a rule discovery task, the 2-4-6 problem first described by Wason. The first experiment instructionally modified subjects confirmatory tendencies. While a disconfirmatory strategy was easily induced, it did not lead to greater efficiency in discovering the rule. The second experiment introduced subjects to the possibility of disconfirmation only after they had developed a strongly held hypothesis through the use of confirmatory evidence. This manipulation also failed to alter the efficiency of rule discovery. In the third experiment, subjects were taught to use multiple hypotheses at each step, in the manner of Platt's “Strong Inference”. This operation actually worsened performance. Finally, in the fourth experiment, the structure of the problem was altered slightly by asking subjects to seek two interrelated rules. A dramatic increase in performance resulted, perhaps because information which in previous tasks was seen as merely erroneous could now be related to an alternative rule. The four studies have broad implications for the psychological study of inference processes in general, and for the study of scientific inference in particular.

Contrast class cues and performance facilitation in a hypothesis-testing task: evidence for an iterative counterfactual model

“…This scoring method has become standard practice in studies with DG task variants from the time it was established by Tweney et al (1980). Table 1 shows the frequency of correct announcements for the DAX rule across the useful and nonuseful contrast class cue conditions for Experiment 1.…”

“…Despite the failure of falsification instructions to improve task performance in the 2-4-6 paradigm, one facilitatory manipulation that has emerged in the literature is to employ "dual-goal" (henceforth, DG) instructions, as introduced by Tweney et al (1980). In this DG version of the task, participants are required to discover two rules that the experimenter has in mind, one called DAX and the other called MED.…”

Hypothesis generation, sparse categories, and the positive test strategy.

“…Although there is evidence that people often do this (e.g., Taplin, 1975;Mynatt, Doherty, & Tweney, 1977;Doherty, Mynatt, Tweney, & Schiavo, 1979), it is not always the case. Many studies explore how entertaining multiple hypotheses affects the process of inference, with some finding that it helps (e.g., Klahr & Dunbar, 1988;Klayman & Ha, 1989), some finding that it does not (e.g., Tweney et al, 1980;Freedman, 1992), and some finding that it depends (e.g., Farris & Revlin, 1989;McDonald, 1990). In view of this, it would be useful to extend the existing formal results to cover the multiple hypothesis case.…”

“…Although it is sometimes difficult to disentangle from the matching bias (Evans, 1972;Evans & Lynch, 1973;Evans, 1998, see also Yama, 2001), the PTS is pervasive. It is observed in rule-learning problems (e.g, Wason, 1960;Taplin, 1975;Tweney et al, 1980;Klayman & Ha, 1989), the four-card selection task (e.g., Wason, 1968;Jones & Sugden, 2001), scientific research (e.g., Mahoney & de Monbruen, 1977;Mynatt, Doherty, & Tweney, 1978;Dunbar, 1993), and many other contexts (Nickerson, 1998). The bias to use a PTS can be ameliorated in some situations (e.g., Johnson-Laird, Legrenzi, & Legrenzi, 1972;Cheng & Holyoak, 1985;Cosmides, 1989), but is rarely completely eliminated.…”

The Role of Anomalous Data in Knowledge Acquisition: A Theoretical Framework and Implications for Science Instruction

“…Psychological research consistently shows that discovering scientific concepts and principles is very difficult. Both children (Dunbar & Klahr, 1989;Forman & Cazden, 1985;Schauble, 1990;Schauble, Klopfer, & Raghavan, 1991) and adults (Dunbar, 1989;Gorman, 1986Gorman, , 1989Tweney et al, 1980) find it difficult to invent hypotheses that adequately cover a domain of data. This is true even of relatively simple domains, where the range of possible hypotheses is both explicitly stated and circumscribed.…”