The Inside versus the Outside View

The Inside versus the Outside View

What are the psychological mechanisms that underlie people’s optimistic forecasts? And how do people segregate their general theories about their predictions (i.e., that they are usually unrealistic) from their specific expectations for an upcoming task? Our analysis begins with an anecdote about a group-authored textbook that was completed years later than expected – each of its authors was aware that such projects typically took years longer than planned, but nonetheless felt confident that the current project would be a pleasant exception. This ill-fated project was in many ways unexceptional, but has become famous as the initial case study for Kahneman and Tversky’s (1979) inside versus outside analysis of the planning fallacy. This analysis builds on a perceptual metaphor of how people view a planned project. In the textbook example, the group of authors focused on the specific qualities of the current task, and seemed to look inside their representation of the developing project to assess its time course and likelihood of completion. The group of authors neglected, however, to look outside the specific project to evaluate the relevant distribution of comparable projects. More broadly, an inside or internal view of a task focuses on singular information: specific aspects of the target task that might lead to longer or shorter completion times. An outside or external view of the task focuses on distributional information: how the current task fits into the set of related tasks. Thus, the two general approaches to prediction differ primarily in whether individuals treat the target task as a unique case or as an instance of a category of similar problems.

For a number of reasons, people favor the inside perspective when developing plans and predictions (Kahneman & Lovallo, 1993; Kahneman & Tversky, 1979). People tend to generate their predictions by considering the unique features of the task at hand and constructing a scenario of their future progress on that task. This is, by and large, what planning means to most people: develop a series of steps that lead from the beginning to a successful conclusion of the project. In deciding how much work to take home for the weekend, for example, individuals may try to imagine when they will start a particular project and how many hours of actual working time they will require. They may also consider the other activities they have planned for the weekend and try to determine precisely when, where, and how they will find time to work on the project. They may even try to envision potential obstacles to completing the project, and how these obstacles will be overcome. Essentially, this internal approach to prediction involves sketching out a scenario that captures how the future project is likely to unfold. Several theorists have similarly noted the tendency to construct scenarios, narratives, or mental simulations as a basis for predicting the future (Dawes, 1988; Griffin, Dunning, & Ross, 1990; Johnson & Sherman, 1990; Kahneman & Lovallo, 1993; Kahneman & Tversky, 1982a; Klayman & Schoemaker, 1993; Ross & Buehler, 2001; Zukier, 1986).

People’s completion estimates are likely to be overly optimistic if their forecasts are based exclusively on plan-based, future scenarios. A problem with the scenario approach is that people generally fail to appreciate the vast number of ways in which the future may unfold (Arkes, Faust, Guilmette, & Hart, 1988; Fischhoff, Slovic, & Lichtenstein, 1978; Hoch, 1985; Shaklee & Fischhoff, 1982). For instance, expert auto mechanics typically consider only a small subset of the possible faults that can occur in a car, and hence underestimate the probability of a breakdown (Fischhoff et al., 1978). Similarly, when individuals forecast personal events they often fail to consider alternatives to their favored scenario (Griffin et al., 1990; Hoch, 1985). Focusing on the target event (the successful completion of a set of plans) may lead a predictor to ignore or underweight the chances that some other event will occur. Indeed, developing a scenario leading to a given outcome seems to interfere with people’s ability to generate scenarios that would yield alternative outcomes (Hoch, 1984; Koehler, 1991). When individuals are asked to predict based on realistic “best guess” scenarios, their forecasts are generally indistinguishable from those generated by “best case” scenarios (Griffin et al., 1990; Newby-Clark, Ross, Buehler, Koehler, & Griffin, 2000).

The act of scenario construction itself may lead people to exaggerate the likelihood of the scenario unfolding as envisioned. Individuals instructed to imagine hypothetical outcomes for events ranging from football games to presidential elections subsequently regard these outcomes as more likely (for reviews, see Johnson & Sherman, 1990; Koehler, 1991). In one study, for example, individuals who explained why they might succeed on an upcoming anagram task predicted that they would perform better than those who explained why they might fail (Sherman, Skov, Hervitz, & Stock, 1981). The implication of this research is straightforward: If people focus almost exclusively on their plans for successful task completion while generating predictions, they may convince themselves that an optimistic forecast is warranted.

One way to reduce the biases associated with scenario thinking (or the inside approach) is to incorporate relevant distributional information into prediction (Kahneman & Tversky, 1979). When individuals make predictions about everyday activities, the distributional information could either be their own past experiences (personal base rates) or the experiences of others (population base rates). The value of attending to both personal base rates (Osberg & Shrauger, 1986; Shrauger & Osberg, 1981) and population base rates (Dunning, Griffin, Milojkovic, & Ross, 1990; Epley & Dunning, 2000; Hoch, 1985; Vallone, Griffin, Lin, & Ross, 1990) has been demonstrated in several studies examining the accuracy and confidence of people’s predictions about their own future behavior. In the Osberg and Shrauger (1986) study, for example, individuals who were instructed to focus on their own previous experiences forecast their activities during the next 2 months (e.g., whether they would skip a class, get high on drugs, change their hairstyle, go to a movie) more accurately than control subjects. Attending to past experience is particularly useful when base rates are extreme; that is, when the behaviors have been either very frequent or very infrequent in the past (Dunning et al., 1990; Shrauger, Mariano, & Walter, 1998; Shrauger, Ram, Greninger, & Mariano, 1996).