Goal-based explanations 1
Running head: Goal-based explanations
Implicit judgments in goal-based explanations
John McClure
Victoria University of Wellington,
Robbie J. Sutton,
University of Keele
Denis J. Hilton
University of Toulouse 2
Paper presented at the Fifth Sydney Symposium of Social Psychology, 2002
This research was supported by a grant from the Science Faculty of Victoria University of Wellington. Correspondence concerning this paper should be addressed to John McClure, School of Psychology, Victoria University of Wellington, New Zealand; email: .
Implicit judgments in goal-based explanations
What processes governs people's explanations of actions? Consider the following simple action: "Mary crossed the quadrangle and bought lunch costing $3 at the local restaurant". People may explain this action in a number of different ways. They may offer explanations such as: "She wanted a meal", or "she wanted to impress friends", which focus on the agent's goals or intentions. Alternatively they may explain the action by saying that: "The café was open", "she had enough money in her wallet", "she was rich", etc. These explanations refer to preconditions, conditions that enables actions to occur. Research shows that in different conditions one of these two types of explanation is usually preferred and the other is discounted, or “left out” of the explanation (Leddo, Abelson, & Gross, 1984). Consider if the scenario above is changed to read: "Mary crossed the street and bought a new Rolls Royce costing $300,000 at the local dealer." Is the preferred explanation different for this scenario, which has a similar grammatical structure to the earlier scenario? And if so, why? And does the choice of explanation relate to the notion of implicit processing, which has been so useful in social cognition?
This chapter outlines factors influencing people's judgments about goal-based explanations and examines whether these judgments about explanations relate to the distinction in social cognition between implicit and explicit processing. The term "implicit" is often used in research on explanations, but its meaning sometimes differs from its dominant use in social cognition. This chapter attempts to clarify connections between the research on goal-based explanations and the concept of implicit processing as it is normally used in research on social cognition.
The concept of implicit processing has been used in two different ways in regard to explanations and social perception. In the first case, implicit processing has referred to the person's cognitive structures and processes through which the person's explanations are constructed from the flow of experience (e.g., Wegner & Vallacher, 1977, 1981). A second more specific use of the concept applies particularly to judgments about actions and traits. Research suggests that behavioural and dispositional inferences involve a two stage process, in which the first (implicit) stage is automatic and involves categorizing a person's behaviour (e.g., Trope & Alfieri, 1997). The second stage involves controlled processing (or an inference) that takes account of situational factors and that may modify or correct the initial automatic inference. It is at this second stage that causal reasoning and causal schemata come into play. This model of dispositional inferences has been applied to the correspondence bias (sometimes called the fundamental attribution error), where observers' spontaneous judgments about actors show a bias toward dispositional attributions that take little account of situational forces (e.g., Gilbert & Malone, 1995). The claim that this tendency exemplifies implicit processing is supported, for example, by research showing that the correspondence bias is accentuated when people are cognitively busy or overloaded. This research demonstrates clear link between dispositional inferences and implicit processing; however, there has been less exploration of the links between other components of the attribution process and implicit processing, although researchers who write about attributions often use the term "implicit" in regard to judgments. This chapter attempts to clarify potential links between implicit processing and explanations, focusing particularly on goal-based explanations.
Research on goals and intentions
In social psychology, most research on explanations for intentional actions suggests that people typically explain these actions in terms of goals. This finding is robust, despite the fact that many actions are influenced by other causes, including environmental factors and internal enabling factors such as abilities. In his seminal work on this issue, Heider (1958) claimed that intentional actions reflect personal causality, the person's intention that leads to the action. This claim is reinforced by the concept of equifinality, which proposes that people circumvent obstacles to fulfill their goals. For example, if you want to go to the restaurant, and find you have no cash, you may use your credit card to circumvent the deficiency and accomplish the goal. In such situations, the goal stays constant and the means to achieve it vary. Because the action is determined more by the goal than the particular means that the actor uses to achieve it, the best explanation of the action is seen in terms of the actors' goal, and this is the explanation that observers are most likely to use. Heider's research on intentions is particularly interesting in regard to implicit processing, because he showed that people spontaneously and "automatically" attributed intentions to geometric shapes on a movie screen, when these shapes moved in sequences that simulated animated action (Heider & Simmel, 1944). This research clarified the parameters of movement that lead to automatic attributions of intention, but most research on attributions after Heider's work shifted focus away from intentional actions, until the issue was resurrected recently by goal-based theories and folk psychology.
The issue of how people judge intentions as explanations is also important in theory about causation and responsibility in legal contexts. Authors dealing with this issue have demonstrated a similar emphasis to Heider. Hart and Honoré (1985), for example, noted that that in court cases many explanations of events refer to both intentional causes and physical causes. For example, in regard to a forest fire, an intentional explanation may be that a person lit the fire, whereas a physical explanation may be that lightning lit the fire. Physical and intentional causes sometimes occur in a sequence, as for example where lightning ignites a bush and a person fans the flames. When an explanation contains both intentional causes and physical causes, Hart and Honoré claimed that people see the intentional causes as a better explanation. According to Hart and Honoré, this preference exists because whereas physical causes are seen as part of the normal order of events, intentions are seen as interventions in the natural order, and hence as more abnormal than physical causes.
The idea that people prefer intentions as explanations is consistent with current research on folk theories, which demarcates people's lay theories of intentional actions. This research shows that intentions are seen as sufficient explanations for actions, whereas other causes are seen as contributory causes rather than sufficient explanations (Malle, 1999, 2001; Malle & Knobe, 1997). For example, Malle and Knobe (1997) examined which features people see as essential to intentional actions, and found that there were two central features: the desire leading to the actions and the belief that the action could be accomplished. In later research, Malle (l999) showed that people think that the accomplishment of an intentional action also requires an awareness of the intention to perform the action, and the minimum skills required for the action. They obtained these results by presenting scenarios with different combinations of these folk psychology elements, and asking people if the intention is present or if the person is likely to perform an intentional action. If people judge that the presence of a cause guarantees the action (in conjunction with other necessary causes) and its absence prevents the action, then that cause is seen as essential to the concept of intention (See also Kashima, McKintyre, & Clifford,1998).
Knowledge structures and goal-based explanations
A second current line of research on explanations of actions is the knowledge structure approach. This avenue draws on artificial intelligence and relates explanations to general cognitive structures that guide people's perceptions and explanations (e.g., Abelson & Lalljee, 1988; Kruglanski, 1996; Lalljee & Abelson, 1983; Read, 1987; Read & Marcus-Newhall, 1993; Schank,& Abelson, 1977; Wilensky, 1983). The knowledge structure approach models people's thinking about scenarios in the context of scripts, rather than examining snapshot explanations. This approach uses the concept of goals rather than intentions, but it assumes that these goals express intentions. Research on knowledge structures recognizes many features that affect actions, but it treats goals as central to both the occurrence and the explanation of actions. Nonetheless it also recognizes the role of preconditions, the circumstances that enable actions to occur, such as having money in the case of hiring a taxi. This taxonomy of goals and preconditions differs from the broad categories of internal and external causes that has predominated in research on attributions, in that some preconditions such as abilities may be internal to the person, rather than external. The knowledge structure approach to explanations provides a useful framework in this chapter for examining the relation between explanations and implicit processing.
A prototypical example of the knowledge structure approach as applied to explanations is seen in Leddo, Abelson and Gross' (1984) seminal studies on explanations of completed and failed actions. These studies present scenarios describing scripted actions that are either completed or not completed, and then present a range of explanations of the action (or non-action). One scenario for example describes John, driving along the freeway, who sees a restaurant up ahead and stops in at the restaurant for a meal. In the non-completed version of the scenario, John does not stop in for a meal. Each scenario is followed by a range of explanations, including a goal (John wanted a meal), two different preconditions (He had enough money for a meal, the restaurant was open), and explanations combining these constituent explanations in conjunctions (e.g., John wanted a meal and he had enough money for a meal). Filler explanations were also listed, to reduce the obviousness of the manipulations.
Participants had to judge the probability of each explanation for the action or non-action. As predicted by the knowledge structure approach, they judged goals to be more probable explanations than preconditions. However, for completed actions they also judged the conjunctions of goals and preconditions as more probable than goals, whereas in the case of non-actions, the absence of either a goal or a precondition was judged as good an explanation as a conjunction of the two causes. This result suggests that people see several causes as necessary for the completion of an action, but see the default of a single key cause as sufficient to prevent the action. This research suggests that for single-cause explanations of actions, people prefer goals over preconditions. However, if given the opportunity, people judge a conjunction of a goal and a relevant precondition as a more probable explanation of completed actions than the goal on its own, whereas with non-actions, the absence of a single goal or precondition is seen as more probable.
Subsequent research has posed two qualifications to these findings. First, research has shown that these findings are limited to causes that facilitate actions and do not generalize to competing or inhibitory causes. When a person does not go to restaurant, their non-action may arise not from the absence of facilitatory causes but from the presence of competing causes, such as competing goals (Wilensky, 1983). Research shows that when people consider competing causes and the relevant preconditions, conjunctions of these causes are judged more probable as explanations for non-actions than for actions (McClure, Lalljee, Jaspars, & Abelson, 1989).
A second qualification to the finding of a general preference for goals as explanations concerns extreme actions. Most of the scenarios in the research described thus far describe relatively common actions such as going to the restaurant or taking a book out of the library. A number of theories suggest that the pattern of explanations obtained with these common tasks may change with extreme or difficult actions. Kelley's (1972) schema model of attribution suggests that for extreme events, people apply a multiple necessary schema and prefer conjunctive explanations that include two or more causes. A related theory developed by Reeder and Brewer (1979) claims that for extreme success, people apply a hierarchically restricted schema, by which they perceive ability as necessary for any substantial achievement. This implies that with extreme success, people are more certain that ability is present than the relevant goal. In support of this view, McClure, Lalljee, and Jaspars (1991) showed that when people explained Einstein's outstanding success as a scientist, they attributed his success to ability, rather than a conjunction of causes as Kelley's theory would suggest (See also Johnson, Boyd, & Magnani, 1994).
To examine the effect of extremity on judgments of goals and preconditions, McClure and Hilton (1997) reframed the scenarios developed by Leddo et al. (1984) to include extreme outcomes, in addition to moderate outcomes. They replicated Leddo et al.'s finding that to explain common actions, participants preferred goals such as going to the restaurant, regardless of whether the actions have been obstructed, and they showed that this preference extends to extreme actions that have not been obstructed, such as rich person buying a very expensive car. For extreme actions that have previously been prevented, however, such as a poor person buying an expensive car, participants judge the acquisition of a relevant enabling condition (such as money) as a more probable explanation than the relevant goal. Subsequent research showed that the preference for goals and preconditions for extreme and common actions is mediated by people's judgments of whether the actions are controllable, rather than the actions' inherent probability (McClure, Densley, Liu, & Allen, 2001).
Different criteria of a good explanation
These findings show that sometimes people prefer goals and at other times they prefer preconditions. Sometimes they prefer single causes such as a goal and at other times they prefer conjunctions. The key question here is what mediates these preferences and whether these mediating factors relate to the distinction between implicit and explicit processing. Most of the initial research on knowledge structures obtained probability judgments about explanations, asking participants to judge the probability of a goal or precondition, given a scenario where the action has occurred. Probability judgments do provide one valuable measures of causal judgments, but there are a number of other criteria by which people can judge explanations, such as necessity, sufficiency, and informativeness. Research suggests that some of these other measures may be better predictors of the perceived quality of an explanation than probability (Hilton & Erb, 1996; McClure, 1998; McGill, 1990, 1991). People can judge the necessity of a cause by rating how likely it is that the action would have occurred if the given cause was absent - obviously low ratings on this counterfactual type of task equate high necessity. Measures can also tap the judged sufficiency of cause, by getting participants to rate the likelihood of the action occurring given that the cause is known to be present. Clearly high ratings on this measure equate high sufficiency.