Availability and Base-Rate Neglect in Expert Advice on Sexual Assault Resistance

As part of a project attempting to derive responsible advice to women for reducing the risk of sexual assault (summarized in Fischhoff, 1992), Marcia Morgan, Lita Furby, and I reviewed a large sample of popular publications. These created a dismaying picture of confident, universal, yet contradictory recommendations. The most striking division was between experts strongly advising all women to undertake physical resistance and those telling women to avoid it. Experts proscribing such resistance were disproportionately from law enforcement sources – rather than, say, rape crisis centers or self-defense classes. Police predominately see women who resisted physically (hence, can better defend themselves in court), women whose resistance failed (meaning that rapes were completed), and women who suffered other injuries (which are easier to prosecute than coerced sexual acts). Thus, police samples are biased toward unsuccessful physical resistance. Police officers would be like other “victims of availability,” if they remembered what they had seen and heard, perhaps in great detail, but lacked an appreciation of what they were not seeing.

If the police worried about such bias, undoing its effects would require understanding its sources, assessing their magnitude, and affording equal cognitive status to unseen events. The last of these steps would confront the difficulty of integrating case-specific and base-rate information, often attributed to reliance on the representativeness heuristic. Base-rate neglect has sometimes been reduced by presenting representative samples case-by-case (e.g., Manis et al., 1980). For police, that would mean meeting women who currently avoid the law-enforcement system. Doing so would require reducing the barriers to seeking police help after sexual assaults.

Judgmental biases can hamper such social change. Assessing the effectiveness of responses to sexual assault is a very difficult task (even for researchers). Overconfident experts have an exaggerated sense of having seen it all. One could hope that experts would systematically look for evidence contradicting their favored beliefs – therefore benefit from the debiasing that such searches produce in lab studies. That search should be particularly valuable when experts second-guess women’s responses to assaults. Misfortunes may be disproportionately salient – hence, induce more hindsight bias – than the “nonevents” of avoiding harm. Unwilling to wait for spontaneous debiasing efforts, advocates for women’s welfare have lobbied for better protection of victims’ rights, dedicated “rape teams,” and community-oriented policing. In addition to their direct value, these reforms create a new balance of evidence, broadening the sample of women coming to the police and the experiences that they are willing to share. Both trends should weaken the overconfidence of simplistic accounts, based on even the most conscientious summaries of biased samples of evidence.

Such explanations can (and, indeed, should) be applied to other advice providers. For example, what do self-defense trainers see vis-à-vis women’s ability to resist sexual assaults physically? Indeed, how do these judgmental processes affect scientists working on the problem? They, too, have favored sources of evidence, and expectations regarding their results. Our project included a secondary analysis of all studies that we could find regarding the effectiveness of self-defense strategies. There were about three dozen, using varying methodologies and analytical procedures. Attempting to avoid some of the judgmental biases in impressionistic summaries (Rosenthal, 1994), we translated all studies into a common analytical framework and unit of effectiveness: the difference in the percentages of completed sexual assaults among women who did and who did not use each strategy. In so doing, we found that several authors had reversed the conditionality in this relationship, looking at the percentage of women who did and did not use a strategy, among those experiencing completed assaults. In effect, they had committed the “base-rate fallacy,” ignoring the percentage using each strategy (Dawes, 1993). Eddy (1982) shows evidence and implications of such confusion among mammographers. We chose to make little note of these biases, in deference to the scientists conducting these very difficult studies.

Our review yielded a moderately consistent signal: Women who resisted physically experienced completed sexual assaults less frequently than women who used other self-defense strategies. That summary fits the preconceptions of some women and practitioners, but not others. For it to make sense to all, it would have to be integrated with their own anecdotal observations and “mental models” of how these processes work (Fischhoff, 1999; Furnham, 1988; Leventhal, Diefenbach, & Leventhal, 1992; Rouse & Morris, 1986). For example, they might be told how assaults can evolve over time and the role of physical resistance in them, as well as why appearances may be deceiving (in the sense of what gets seen and reported). Such debiasing falls into the general category of improving the match between individual and task. A specific category (not mentioned explicitly in Table 41.1) might be “integrating episodic and semantic beliefs.” Doing so is part of the training in many professional fields, whose apprenticeship includes working through specific examples in the context of general theories.

This case offers a plausible application of heuristic theories, rather than a conclusive test of them. The account would be stronger if (1) bridging theories specified where and how the heuristics should operate and (2) similar conditions produced bias in laboratory and real-world settings. Surveys of advice givers’ beliefs and information sources could test these proposed explanations. Their results would need to be assessed alongside the broader evidence on heuristic thinking.