What Makes an Explanation Believable?: Mechanistic and Anthropomorphic Explanations of Natural Phenomena

What Makes an Explanation Believable?: Mechanistic and Anthropomorphic Explanations of Natural Phenomena Jordan Schoenherr (psychophysics.lab@gmail.com) Department of Psychology, Carleton University 1125 Colonel By Drive, Ottawa, ON K1S5B6 Canada Robert Thomson (rthomson@connect.carleton.ca), Jim Davies (jim@jimdavies.org) Institute of Cognitive Science, Carleton University 1125 Colonel By Drive, Ottawa, ON K1S5B6 Canada participants are given premises and are required to indicate whether a conclusion logically follows from them. To examine the effects of prior beliefs on successful task completion, Evans et al. (1983) varied both the validity of the argument and believability of the conclusion. Validity follows from rules of formal logic whereas believability stems from how closely the conclusion conforms to one’s prior beliefs. When the argument is invalid but the conclusion is believable, participants should disregard their prior belief and focus on the invalidity of the argument. A belief bias is observed when participants accept these invalid yet believable arguments. Abstract Many biases in decision-making and reasoning are a result of ignoring logical rules and relevant information while focusing on irrelevant cues present within an argument. In the present study we examine explanatory schemata – a set of interrelated concepts - that are deemed relevant to participants. Participants were first trained in a syllogistic reasoning task and were then presented descriptions of natural phenomena and explanations. An instructional manipulation varied the source of the explanations (scientists or people) as well as the animacy of the natural phenomena (living or nonliving). Explanations used either mechanistic (e.g., force) or anthropomorphic (e.g., wants) terms. We found that participants were more accurate when assessing mechanistic explanations. Keywords: anthropomorphism; reasoning; belief bias heuristics; Knowledge Effects in Decision-Making Failures of decision-making have also been observed in the Wason Selection Task (Wason, 1966). In this task, participants are presented with four cards in order to identify whether a rule is false (e.g., If a card has a vowel on the obverse it will have an even number on the reverse). To successfully complete the task, participants should select a card that would disconfirm the rule (an odd number) and one that confirms the statement (a vowel). Over a wide range of subject categories, average responses rarely exceed 25% accuracy (Cosmides & Tooby, 1992; Stanovich, 2004). Performance in the Wason Selection Task can be improved when domain-specific knowledge facilitates the selection of an accurate response (Cosmides & Tooby, 1992; Gigerenzer & Selten, 2000). To account for this evidence, Cosmides (1985) speculated that failures in reasoning tasks could be attributable to a mismatch between the domains considered in the task and domain-specific cognitive modules created through natural selection. If these tasks reflected verification of violations of social contracts – served by an innate cognitive module in her account – then participants’ performance should improve (Cosmides 1989; Cosmides & Tooby, 1992). For instance, participants could be presented with the task of verifying that customers of a pub are of an appropriate age for entry. Studies that have employed such methodologies have observed performance at or exceeding 75% accuracy (e.g., Gigerenzer & Hug, 1992; Griggs & Cox, 1982). Thus, if prior knowledge is available, the extent to which it overlaps with task demands should determine performance (cf. Liberman & Klar, 1996). Given that people possess both naive psychological and physical theories about the world – whether learned or syllogistic Introduction Each day, the media presents information to the public from purportedly credible sources. People must then generate beliefs based on these explanations. This is especially true of scientific discoveries. Weisberg, Keil, Goodstein, Rawson and Gray (2008) found that explanations from the psychological sciences were seen as more satisfying when accompanied by irrelevant neuroscientific information. This result is also supported by prior research, which has identified that the kind of explanation (Cosmides & Tooby, 1992) and prior knowledge (Kahneman & Tversky, 1982) affects the accuracy of people’s judgments. In the present study we examined the kinds of prior knowledge that can lead to inaccurate assessments of explanations of natural phenomena by manipulating the source of the explanation, the properties of the natural phenomena and the mode of explanation. We additionally used subjective confidence reports to determine whether participants were aware of the factors influencing their performance. Biases from Domain-Specific Knowledge Subjective biases are generally attributed to a variety of decision-making rules and heuristics (for a review see Kahneman, 2003; Kahneman & Tversky 1982). Heuristic- related biases are also observed in the context of rule-based syllogistic reasoning tasks (e.g., Evans, Barston & Pollard, 1983; Sa et al., 1999). In a typical syllogistic reasoning task,