Academic journal article The Psychological Record

Categorical Induction as Hypothesis Assessment

Academic journal article The Psychological Record

Categorical Induction as Hypothesis Assessment

Article excerpt

The act of generalizing the properties of a previously observed thing to a novel thing or to a whole category of things is quite pervasive in our day-to-day lives. We frequently generalize properties from one instance of a category to another instance of the same category (e.g., "My new collie will probably have the same funny eating habits that my first collie had."); from instances to categories (e.g., "My boyfriend drives me crazy with his take-charge attitude. I think that all men must be endowed with some kind of domination impulse."); and from one category to another (e.g., "These new Russian immigrants will probably be just as clannish as the Chinese immigrants are.").

Arguments involving natural biological categories have proven to be an especially useful tool for studying inductive generalizations (e.g., Collins & Michalski, 1989; Lopez, 1995; McDonald, Samuels, & Rispoli, 1996; Osherson, Smith, & Shafir, 1986; Osherson, Smith, Wilkie, Lopez, & Shafir, 1990; Rips, 1975; Sloman, 1993). Keil (1989) has argued that biological categories possess an exceptionally rich network of property intercorrelations, and this makes them especially well suited for studies of property generalizations. In addition, the clear hierarchical structure of biological categories allows for several distinct argument forms, including general-conclusion arguments in which the premises are subsumed by the conclusion (e.g., (All Robins Have Substance X, All Bluejays Have Substance X [therefore] All Birds Have Substance X}), and specific-conclusion arguments in which the conclusion is at the same level of classification as the premises (e.g., [All Robins Have Substance X, All Bluejays Have Substance X [therefore] All Ge ese Have Substance X}). The simple, formal structure of such arguments also allows for the ready addition, deletion, or substitution of categories; and for the use of either neutral predicates, such as Have Substance X, or familiar predicates, such as Eat Meat. Because arguments with neutral predicates are referred to frequently throughout this paper, they will hereafter be summarized by omitting both the adjective and the predicate (e.g., {Robins, Bluejays [therefore] Geese}).

McDonald et al. (1996) successfully accounted for most of the variability in the perceived strength of a set of general-conclusion arguments by employing a hypothesis-assessment model that treats argument strength estimation as a form of hypothesis assessment. This model postulates predictors that are already well known to affect perceived hypothesis plausibility across a wide range of reasoning tasks, thus avoiding the need to postulate factors unique to the phenomenon under investigation. The two primary goals of the present paper are to describe the theoretical constructs by which the proposed model explains specific-conclusion argument strength, and to test empirically the model's ability to account for variability in the perceived strength of specific-conclusion arguments.

The hypothesis-assessment model treats the premises of an inductive categorical argument as evidence and the conclusion as a hypothesis. For example, {Penguins [therefore] Birds} is treated as an assertion that the empirical observation that all Penguins possess a certain property makes it reasonable to hypothesize that all Birds might possess this same property. The model attempts to account for the varying strength of categorical arguments by postulating four factors that are said to directly affect the perceived plausibility of hypotheses in general: (a) the logical consistency of the hypothesis with the evidence, (b) the number of accessible alternatives to the hypothesis, (c) the accessibility of the target hypothesis itself, and (d) the scope of any required generalization. These factors are well established as being relevant to the perceived plausibility of hypotheses in less abstract hypothesis-testing and assessment tasks (references provided in the discussion immediately below), and the proposed model merely generalizes these factors to the phenomenon of categorical argument strength. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.