When we look back on what we knew previously, we are often biased by what we know now. In hindsight, we think that we “knew it all along” (Wood, 1978), we assign higher a priori probabilities to facts or outcomes (Fischhoff, 1975), and we misremember our prior predictions as closer to facts or actual outcomes (e.g., Erdfelder & Buchner, 1998). This cognitive illusion has been termed hindsight bias (for reviews, see Blank et al., 2007; Roese & Vohs, 2012) and is pervasive in a variety of everyday situations, such as in political elections, medical diagnoses, or scientific experiments (e.g., Arkes et al., 1981; Blank et al., 2003; Slovic & Fischhoff, 1977). It is also robust across a wide range of materials and different tasks (Pohl, 2007).
In one of the most common types of task to study hindsight bias, the memory paradigm, participants are first asked to provide judgments; for example, numerical judgments to a set of difficult knowledge questions (e.g., How many letters does the Arabic alphabet contain?). They then receive the correct answer for some or all of the questions (e.g., The Arabic alphabet contains 28 letters). Finally, they are asked to recall their own original judgments. Hindsight bias occurs when recalled judgments are shifted toward presented correct answers. We will refer to the bias measured in the memory paradigm as the memory component of hindsight bias.Footnote 1
Research has repeatedly demonstrated that in such numerical-estimation tasks, older adults are more prone to the described memory distortion (e.g., Bayen et al., 2006; Bernstein et al., 2011; Coolin et al., 2015; Groß & Bayen, 2015). In a recent meta-analysis, Groß and Pachur (2019) found that older adults show worse memory for their own original judgments than young adults do, and—as a consequence—must reconstruct their judgments more often. In this reconstruction, older adults are more prone to be biased by the correct answer than younger adults are. This could be due to older adults’ worse memory for their own original judgments (e.g., Groß & Bayen, 2015), and/or to their larger difficulties in inhibiting the correct answer in the reconstruction of original judgments (e.g., Coolin et al., 2015).
In this study, we aimed at extending previous findings on age differences in hindsight bias with regard to two important aspects. Our first main aim was to investigate age differences in hindsight bias in a self-relevant domain with positive and negative outcomes. Our second main aim was to expand our understanding of age differences in hindsight bias by investigating manifestations other than memory distortions—namely, retrospective impressions of foreseeability and inevitability. In the following, we will explain in more detail the importance of these two extensions in research on age differences in hindsight bias.
Hindsight bias for positive and negative outcomes
The finding that older adults are more prone to hindsight bias than younger adults are is based primarily on studies that investigated memory for numerical judgments in general-knowledge tasks (summarized in Groß & Pachur, 2019).Footnote 2 However, hindsight bias not only occurs after learning facts and figures that are emotionally neutral, but also after emotionally significant outcomes of personal relevance, such as losing a job, missing a sale, or hearing from an old friend (Blank & Peters, 2010; Louie, 1999; Mark & Mellor, 1991; Pezzo & Beckstead, 2008; Tykocinski, 2001). For example, in one of the first studies involving emotionally relevant outcomes, Mark and Mellor (1991) found hindsight bias to be reduced in laid-off manufacturing workers, for whom the layoff was negative, compared with retained workers or community members, for whom the layoff was less consequential. Subsequent studies have found similar decreases in hindsight bias after negative outcomes (Groß et al., 2017; Nestler et al., 2010), but other studies have also found increases in hindsight bias after negative, self-relevant outcomes (e.g., Tykocinski, 2001; Wann et al., 2008). Thus, while the valence of the outcome does affect hindsight bias, this effect is not uniform. As we will explain in more detail in the next section, among the key variables that moderate the effect of outcome valence on hindsight bias in self-relevant domains are the amount of personal controllability of the (positive or negative) outcome (Blank & Peters, 2010), and the hindsight-bias manifestation (component) under consideration (Blank et al., 2008).
To the best of our knowledge, no study has examined age differences in hindsight bias for such event outcomes. There is reason to believe that positive and negative outcomes—that is, outcomes that can be emotionally significant—have a special status in older adults’ cognitive processing. It is thought that older adults shift their goals towards emotional functioning and meaningfulness as their future time perspective becomes less expanded (Carstensen, 2006). In fact, older adults report higher levels of emotional well-being than younger adults (e.g., Carstensen et al., 2011; Kunzmann et al., 2017; Mroczek & Spiro, 2005). Prominent theories of emotional aging postulate that these age-related differences in affective experience can be linked to age-related differences in cognition. Specifically, in comparison with young adults, older adults show a preference for positive over negative material in information processing, including memory (Mather & Carstensen, 2005; Reed et al., 2014). For example, compared with younger adults, older adults tend to recall fewer negative than positive images (Charles et al., 2003), and they recall more positive and less negative affect than do younger adults (Levine & Bluck, 1997; Ready et al., 2007). Age-related differences in cognition may thus be in service of emotion regulation (for a review, see Isaacowitz & Blanchard-Fields, 2012). We assume that this may also show in the processing of positive and negative outcome information in hindsight judgments. Therefore, a first main aim of the current study was to investigate age differences in hindsight bias after positive and negative outcomes.
Multiple components of hindsight bias
As outlined above, hindsight bias can manifest itself as biased recall of prior given judgments, once the facts or outcomes are known, typically referred to as the memory component of hindsight bias. However, knowing about facts or outcomes of events also affects retrospective judgments of inevitability (“It was bound to happen”; Fischhoff, 1975), and retrospective judgments of foreseeability (“I knew it all along”; Fischhoff, 1977). To accommodate the different manifestations of hindsight bias, Blank et al. (2008; see also Roese & Vohs, 2012) proposed a multiple-components view on hindsight bias. According to this view, the three components—the memory component of hindsight bias, inevitability impressions, and foreseeability impressions—are independent manifestations of hindsight bias that can occur individually or in combination in reference to an event’s outcome, depending on the requirements of the task. That is, they can be empirically dissociated (e.g., Nestler et al., 2010).
The memory component of hindsight bias is mainly governed by variables that affect memory in general, such as the depth of encoding, the length of the retention interval, and the amount of interference (Erdfelder et al., 2007; Groß & Bayen, 2015; Hell et al., 1988). It has been suggested that this memory component of hindsight bias is a by-product of an adaptive knowledge-updating process (Hoffrage et al., 2000).
Inevitability impressions are a second component of hindsight bias. They occur when one retrospectively thinks “it had to happen,” or when—with outcome knowledge; that is, in hindsight—people assign higher probabilities to outcomes than without outcome knowledge (i.e., in foresight; e.g., Fischhoff, 1975). For example, in a study by Slovic and Fischhoff (1977), participants were asked to rate the probabilities of outcomes of scientific studies (e.g., a female rat showing maternal behavior after having been injected with blood from a mother rat). Participants who had been informed about an alleged true outcome (hindsight group) assigned higher probabilities to these outcomes (and the outcomes’ replicability), compared with participants who had not been informed about a true outcome (foresight group). Inevitability impressions follow from a person’s causal model; therefore, they are stronger when meaningful explanations for an outcome are available, compared with when no explanations are available (Nestler et al., 2010).
Foreseeability impressions are the third component of hindsight bias and refer to the metacognitive belief that one was able to foresee or predict the occurrence of the outcome (“knew-it-all-along” effect). In contrast to inevitability impressions, which refer to the probability of the situation or outcome itself (i.e., Was the outcome inevitable?), foreseeability impressions are judgments about one’s thinking about the outcome (i.e., Did I foresee the outcome?). For example, participants in a study by Hastie et al. (1999) reviewed information about a route of hazardous train tracks. Participants who had learned about the occurrence of a severe accident (hindsight group) indicated that they had foreseen the accident to a greater extent than did those who had not learned about the accident (foresight group). Foreseeability impressions are counter to feelings of surprise (Nestler et al., 2010).
Inevitability impressions and foreseeability impressions are not only different in nature, but they might also serve different psychological functions. Specifically, it has been suggested that individuals may adjust their impressions of inevitability or foreseeability of event outcomes in retrospect in order to regulate negative affect (e.g., Blank et al., 2008; Pezzo & Pezzo, 2007). For example, a person could increase the perceived inevitability of a negative outcome (“It was bound to happen anyway”) in order to cope with disappointment. Tykocinski (2001) found that participants who experienced a disappointing outcome (e.g., the failure to secure an attractive deal) retrospectively showed increased impressions of inevitability of the outcome. Such inevitability increases should be helpful primarily when the outcome had been out of one’s personal control (Blank & Peters, 2010).
To avoid regret or self-blame, a person may decrease the perceived foreseeability of a negative event outcome (“I didn’t see it coming”; Louie, 1999). Such unforeseeability impressions can reduce self-blame for outcomes that had been, at least partly, under the person’s control (Blank & Peters, 2010) and are typically found in contexts with emotional relevance (e.g., Groß et al., 2017; Mark & Mellor, 1991; Nestler et al., 2010). For example, Nestler et al. (2010) found that participants who had experienced a (fictitious) financial loss due to a risky stock purchase rated their loss as less foreseeable in hindsight than in foresight.
Studies that attempted to assess such psychological functions of inevitability and foreseeability impressions, for instance, by examining associations between the hindsight components and affect, have mainly relied on hypothetical scenarios. The results are mixed and inconclusive (e.g., Blank & Peters, 2010; Groß et al., 2017; Tykocinski & Steinberg, 2005). This may indicate that a function of hindsight bias may not be present under all circumstances (discussed by Nestler et al., 2010) or that such a function may be difficult to tap (discussed by Groß et al., 2017).
In sum, it is currently unknown whether age differences in hindsight bias are restricted to the memory component, or whether age differences also emerge for inevitability and foreseeability impressions. A psychological function might be associated with these components of hindsight bias, although empirical evidence is inconclusive. Without knowledge of possible age differences in all three components of hindsight bias, the picture of hindsight bias in older age is incomplete. Therefore, the second main aim of our current study was to fill a research gap by investigating possible age differences in all three components of hindsight bias.
The present study
To address our two main aims, we examined different manifestations of hindsight bias after positive and negative outcomes of everyday-life scenarios. Healthy older and younger adults listened (via headphones) to a series of everyday-life scenarios that had a positive or a negative outcome. To investigate hindsight bias (i.e., a bias due to outcome knowledge), we had participants rate each hindsight-bias component from both a pre- and a postoutcome perspective.
Apart from our two main aims, we explored two additional questions. First, we included cognitive (recall ability) and motivational (future time perspective, current mood) variables to explore potential associations of these variables with hindsight bias and age differences therein. Second, we explored the relationship of hindsight bias (and age-differences therein) with affective reactions to gauge potential age differences in the assumed psychological functions. To this end, we had older and younger adults rate the affective reaction they thought they would experience after the imagined positive and negative outcomes.
With regard to our first main aim, we expected a stronger memory component of hindsight bias in older compared with younger adults (Groß & Pachur, 2019). In line with previous findings on age differences in emotional information processing (Charles et al., 2003), we expected older adults to (a) hold more positive expectations about the outcomes than younger adults, and (b) recall more scenarios that ended positively rather than negatively, compared with younger adults. With regard to age differences in the memory component of hindsight bias for negative versus positive outcomes, our expectations were less clear-cut. Assuming that memory for expectations followed by negative (vs. positive) outcomes would be worse in older than in younger adults due to an age-related positivity effect in recall, older adults’ hindsight bias should be more pronounced for negative (vs. positive) outcomes, because hindsight bias increases as memory for the original judgment decreases (Groß & Bayen, 2015). However, prior research has also found lower emotional well-being, as indicated by higher levels of depressive symptoms, to be associated with stronger memory hindsight bias for negative (vs. positive) outcomes (Groß et al., 2017). Thus, older adults, whose emotional well-being is superior, might show less pronounced memory hindsight bias for negative outcomes than younger adults. Therefore, age differences in the memory component of hindsight bias for negative versus positive outcomes may depend on the relative influence of cognitive (recall) versus emotional-motivational factors in remembering the initial expectation.
If foreseeability and inevitability impressions have psychological (self-regulation) functions that unfold in reaction to the scenarios, then we should observe pre-to-post outcome increases in inevitability and/or decreases in foreseeability following negative outcomes, whereas for positive outcomes, pre- to postoutcome shifts should be less pronounced or in the opposite direction. This pattern should be more pronounced in older than in younger adults if hindsight bias has a psychological function relevant to emotional well-being, which is superior in older adults.