The process of inductive inference in groups: The use of positive and negative hypothesis and target testing in sequential rule-discovery tasks.

Abstract

Two experiments on the process of rule discovery in groups were conducted using a card deck and Wason number triple tasks. The positive effects of incentives on the proportion of correct hypotheses could not be explained by differential testing. The variables "hypothesis vs. target testing" and "small vs. broad range of rule" did not affect this proportion. There was a preponderance of positive tests, declining during trials. Negative tests occurred more often under the target test and broad rule range conditions. The positivity heuristic diminished in importance during trials, whereas the sufficiency, necessity, and Klayman and Ha heuristics increased. Groups produced fewer false and more correct results than individuals. According to the probabilistic model of opinion change, plausibility and faction size contributed to the group process differentially, depending on the task.

Effects of amount of evidence and range of rule on the use of hypothesis and target tests by groups in rule-discovery tasks

“…But they can draw conclusions from the feedback they get after positive or negative tests. Crott et al (1998), McDonald (1990), and Penner and Klahr (1996) found that participants were in fact able to react differentially and adequately to these two relations. We expect that this will also be the case in the present experiment:…”

“…Crott, Giesel, and Hoffmann (1998), Klayman and Ha (1989), Laughlin, Magley, and Shupe (1997), McDonald (1990), Penner and Klahr (1996), and Slowiaczek, Klayman, Sherman, and Skov (1992) found that participants reacted to certain experimental variations in a manner corresponding to Klayman and Ha's conclusions. These findings will be referred to in the following sections.…”

“…Therefore, in this case, negative testing is more appropriate to promote falsifications. 2 Crott et al (1998) studied the effect of the range of the rule empirically. In this study, participants were informed whether the range of rule was small, p(t) 5 .50, or large, p(t) 4 .50.…”

“…Despite the different effects of Htests and Ttests, almost all studies on rule discovery are run with an Htest paradigm. Crott et al (1998) first analysed the use of Htests and Ttests. In contrast to the present experiment, they studied the two kinds of tests under separate experimental conditions.…”

“…1;McGlynn, Tubbs, & Holzhausen, 1995). A third class of experiments considers the use of members' resources within the group (e.g., Crott et al, 1998;Laughlin et al, 1995, Exp. 3).…”