Dual Process Models of Reasoning From the Adult Literature

The study of heuristics and biases in adults has led to several important insights about human reasoning. First, typically developing individuals display systematic errors in their thinking that are attributable to the “design of the machinery” as opposed to the “corruption of thought by emotion” (Kahneman, 2011). One of the important characteristics of many heuristics and biases tasks is that there is a conflict between two types of process that involve different levels of deliberation and awareness, Type 1 and Type 2. The defining feature of Type 1 processes is autonomy, that these processes are necessarily executed in the presence of triggering stimuli (Stanovich, 1999, 2009a, 2011; Stanovich & Toplak, 2012). These processes also tend not to put a heavy load on central processing capacities. Examples of these processes include behavioral and emotional regulation, encapsulated modules for solving specific adaptive problems that have been suggested by evolutionary psychologists, and overlearned associations. Alternatively, Type 2 processes operate serially, and are generally slower and more computationally expensive. It is important to note that Type 1 and Type 2 processes may each lead to correct or incorrect responses (Evans & Stanovich, 2013). However, the design of several heuristics and biases tasks is to create a conflict between Type 1 and Type 2 processes, where Type 1 processes lead to an incorrect response and Type 2 processes lead to a correct response. In these instances, the purpose of Type 2 processes is to compute a better response when the response derived from Type 1 processes needs to be overridden. Type 1 processes often provide input for optimal responses in typically experienced situations, but in unique and novel situations, Type 1 processes only provide a ballpark estimate, requiring the hypothetical simulation of analytic processes to compute a better response (Stanovich, 2011). In order to engage Type 2 processes, the capacity to interrupt and suppress the Type 1 processes must be available to engage processes of hypothetical reasoning. Optimal performance on several of these tasks requires successful override of Type 1 processes in order to avoid a miserly, incorrect response. It is important to note that some participants may not experience this conflict and are able to immediately generate a correct normative response (Bago & De Neys, 2017) if they have consolidated the knowledge that becomes autonomously activated when performing these tasks (Stanovich, 2018; Stanovich et al., 2016).

It is the conflict between Type 1 and Type 2 processes that makes these heuristics and biases tasks particularly difficult for many participants, and it is this distinction between these two types of responses that has also formed the basis for modern dual process models of reasoning (Evans, 2003, 2008, 2010; Evans & Stanovich, 2013; Stanovich, 2009a, 2011). The manner in which these two sets of processes or systems interact has been discussed extensively (Evans, 2003, 2008, 2010; Evans & Stanovich, 2013; Stanovich, 2009a, 2011; Stanovich & West, 2008b). Type 1 and Type 2 processes have also been referred to as System 1/System 2 processes (Evans, 2008; Evans & Stanovich, 2013) and heuristic and analytic processes (e.g., Klaczynski, 2001), the latter most often used in the developmental literature.

The findings from the heuristics and biases literature on adults have consistently shown that many adults tend to demonstrate less than rational responding, often attributable to failing to recognize when an alternative response may be needed and/or a failure to override a dominant response generated by a Type 1 process (Stanovich, 2009a, 2011; Stanovich & West, 2008b). The distinction between Type 1 and Type 2 processes in the adult literature provides a useful way to characterize how different responses occur in these tasks.

Given historical perspectives on the development of reasoning from Piagetian models and advances in the understanding of adult reasoning based on the heuristics and biases tradition, there is much work to be done to further articulate and understand how these processes originate and develop in child and youth samples. Importantly, in order to advance a developmental taxonomy for understanding judgment and decision-making performance, several issues will need to be taken into consideration: (a) the stimulus equivalence problem; (b) confusions with terminology and predictions from dual process models; and (c) the use of multiple indicators of cognitive sophistication.

The Stimulus Equivalence Problem

The heuristics and biases literature with adults has been hugely influential for informing models of rational thinking; however, the stimuli that have been used in adult studies have perhaps created considerable complexity for developmentalists. In particular, there are important issues with respect to the acquisition and consolidation of knowledge required for these tasks and whether child participants experience the conflict between competing Type 1 and Type 2 processes (as has been described in adult samples).

Several judgment and decision-making tasks require specific knowledge, such as numeracy, probabilistic and statistical thinking, and scientific thinking (Stanovich et al., 2016). For example, individual differences in numeracy have been found to predict performance on rational thinking and decision-making in adults (Peters, 2012; Peters et al., 2006; Sinayev & Peters, 2015; Weller et al., 2013). It has been suggested that even before children enter school, they demonstrate a functional understanding of probability (Schlottmann & Wilkening, 2011). In particular, children as young as four years of age may be able to represent the certainty and uncertainty of events (Byrnes & Beilin, 1991). However, several judgment and decision-making tasks require understanding of more complex types of knowledge. For example, demonstration of the conjunction effect presumes an understanding that the probability of the intersection of two elements is statistically less likely than the probability of a single element. Therefore the degree and complexity of specific knowledge required for several heuristics and biases tasks vary, which will have implications for whether certain tasks are suitable and measurable in developmental samples.

As previously described, an important feature of several judgment and decision-making tasks is the conflict between Type 1 processes that may lead to an incorrect response and Type 2 processes that may lead to a correct response. Resolving conflict of competing responses can be illustrated with the classic Stroop test. A sample of a typical Stroop task is presented in Table 1. In Part 1, participants are asked to name the color of each rectangle. In Part 2, participants are asked to read the words (which name the colors). In Part 3, participants are asked to name the color of each word, which in each case is printed in a different color font than the name of the word (interference condition). The interference condition is the hardest condition, as it is difficult for participants to name the color of the font when they have well-developed reading skills. That is, participants must resist reading the word in order to retrieve the name of the font color of the word. Since children have well-developed basic reading skills by around ages eight to nine, or grade three (Chall, 1983), children will likely experience a conflict or interference between naming the colors and reading the words. The reasons that the Stroop task is usable and measurable across children and adults is because across these developmental periods, both groups will experience the conflict that they must override in order to correctly complete this task.

The same issue of experiencing and overriding conflict as in the Stroop task is also relevant for the measurement of judgment and decision-making tasks in children and adults. The design of several heuristics and biases tasks is such that there is meant to be a conflict between a response that is easily elicited by Type 1 processes (that will lead to an incorrect response) and the opportunity to generate a less easily elicited response that can be generated by Type 2 processes (that will lead to a correct response). That easily elicited Type 1 process also requires some overlearned or consolidated knowledge, and whether children in developmental samples have acquired and overlearned that knowledge on some of these tasks is unclear. In addition, the design and administration of the Stroop test makes the conflict between these responses very salient. During the administration of the Stroop task, it is obvious that the participants respond more slowly in the interference condition than in the other conditions. Moreover, participants often make errors that demonstrate that this conflict has occurred, such as accidentally reading the word instead of naming the color, but then quickly self-correcting their response. All of these observations demonstrate the conflict created by the incongruent nature of the interference condition. The presence of this type of conflict is not apparent on judgment and decision-making tasks. In addition, the examiner does not signal to the participant in the instructions as to what is optimal performance on the judgment and decision-making tasks. Even in adult samples, inferences are made regarding whether there has been a conflict between different processes on these tasks, but it is even less clear when such conflict arises in developmental samples. Understanding the development of Type 1 and Type 2 processes is fundamental for properly characterizing different responses on heuristics and biases tasks in child and youth samples (Stanovich, West, & Toplak, 2011b).

Translating Predictions From Dual Process Models in the Adult Literature: Avoiding the Confusions

Some of the confusions that occur in the developmental literature are parallel to those that have occurred in the adult literature. Evans (2008; Evans & Stanovich, 2013) has pointed out that analytic processes (or Type 2 processes) have mistakenly been equated with normatively correct responding and that heuristic processing (Type 1 processes) has been mistakenly equated with non-normative incorrect responding. Instead, both types of processes can often lead to normative correct responding. It is just statistically more likely for Type 2 processes to provide the correct response on the types of heuristics and biases tasks that have been under study because they are designed to elicit Type 1 processes that lead to incorrect responding (Stanovich et al., 2011b). Individual differences in knowledge may also explain findings that have shown that some participants may initially and easily generate a correct response (Bago & De Neys, 2017) if they have overlearned the knowledge required for the task.

These confusions between processes and responses have also pervaded the developmental literature. In particular, this has been perpetuated by the use of the terms “analytic” (Type 2) and “heuristic” (Type 1) responding, instead of characterizing these as analytic and heuristic processes that may lead to correct/normative or incorrect/non-normative responses. Not all Type 1 processes lead to biased, incorrect responses and not all Type 2 processes lead to correct responses (Evans, 2012; Evans & Stanovich, 2013; Stanovich & West, 2008b). However, on heuristics and biases tasks, these tasks are designed to produce conflicting responses and Type 2 engagement is needed to override the response that is championed by Type 1 processes in order to derive a normatively correct response. Both Type 1 and Type 2 processes develop during childhood. Type 2 processes do not replace Type 1 processes and Type 1 processes do not replace Type 2 processes (Brainerd & Reyna, 2001; Markovits, 2013a; Stanovich et al., 2011b), an issue that has been referred to as the “illusion of replacement” (Brainerd & Reyna, 2001).

More Than Age Differences: Using Multiple Indicators of Cognitive Sophistication

In some developmental work, the concept of cognitive sophistication has been used to generate testable hypotheses regarding the development of judgment and decision-making performance (Toplak, West, & Stanovich, 2014b). Specifically, cognitive sophistication poses the following general question: “Do individuals with more complex cognition tend to give a disproportionate amount of normative responses on reasoning tasks?” (Stanovich, 2011). Specifically, individuals who fully understand the normative issues presented in a decision will be more likely to accept the correct normative model, also called the understanding/acceptance assumption. In addition, more cognitively sophisticated individuals will be more likely to derive a correct normative response. That is, individuals with increased cognitive abilities and dispositional tendencies toward open-mindedness and persistence in thinking will be more likely to generate normative responses on these tasks.

Developmentally, cognitive sophistication can be empirically examined in three ways: (a) developmental age differences, (b) correlations with cognitive abilities, and (c) correlations with dispositional tendencies toward open-mindedness and persistence in thinking (Kokis et al., 2002; Toplak et al., 2014b). Developmental differences have been assessed by comparing performance on judgment and decision-making tasks in different age groups of children and youth. Cognitive abilities include the assessment of accuracy and processing efficiency on highly structured tasks where the goals and objectives are provided by the examiner (Stanovich, 2009b), such as measures of intelligence and executive function tasks (Miyake & Friedman, 2012; Reyna, 1995; Salthouse, Atkinson, & Berish, 2003). Thinking dispositional tendencies, or cognitive styles, refer to metacognitive strategies that also support optimal choices on judgment and decision-making tasks (Sinatra & Pintrich, 2003; Stanovich, 2009a, 2011; Sternberg, 2003). For example, the tendency to evaluate beliefs based on new evidence (actively open-minded thinking), persistence in considering alternatives (need for cognition or deliberative thinking), and consideration of future consequences (or future orientation) have been positively correlated with optimal performance on judgment and decision-making tasks in adults (Cacioppo, Petty, Feinstein, & Jarvis, 1996; Stanovich, 2011; Stanovich & West, 1997; Stanovich et al., 2016; Strathman, Gleicher, Boninger, & Edwards, 1994). Overall, it has been reported that increased cognitive sophistication also tends to lead to more optimal responding on judgment and decision-making tasks in developmental samples (Kokis et al., 2002; Toplak et al., 2014b). That is, older children and youth tend to perform better on judgment and decision-making tasks than younger children. Individuals who display higher cognitive abilities tend to perform better on judgment and decision-making tasks. Thinking tendencies that promote evaluation and persisting in thinking will also tend to improve performance on judgment and decision-making tasks. These different indicators of cognitive sophistication provide separable empirical tests and strategies to examine convergence and consistencies in understanding the development of judgment and decision-making performance. Exclusive reliance on any one of these indicators makes it challenging to understand and make sense of unexpected findings, such as developmental reversals (Weldon, Corbin, & Reyna, 2014).

Resistance to Miserly Information Processing

Several of the judgment and decision-making tasks in the literature vary in the extent to which successful performance requires overriding miserly information processing (Stanovich, 2009a, 2011; Stanovich, West, & Toplak, 2011a). Of the tasks where performance is primarily determined by miserly information processing, one of the most well-studied paradigms in the developmental literature has been syllogisms with belief bias content. Belief bias syllogisms represent a special category of deductive conditional reasoning where the believability of the conclusion conflicts with the logical structure of the syllogism, which are typically called inconsistent or conflict problems. Children require basic competence in conditional reasoning skills to successfully solve consistent or no conflict problems. There is an extensive literature on the age-related development of conditional reasoning skills on concrete and abstract problems that are consistent or have no conflict (Markovits, 2013b, 2018; Markovits & Lortie-Forgues, 2011; Markovits & Thompson, 2008; Markovits & Vachon, 1990; Markovits et al., 1996). In studies that have examined performance on belief bias syllogisms in developmental samples, there is a fair amount of convergence to suggest that older children tend to outperform younger children (De Neys & van Gelder, 2009; Evans & Perry, 1995; Handley, Capon, Beveridge, Dennis, & Evans, 2004; Kokis et al., 2002; Markovits & Bouffard-Bouchard, 1992; Steegen & De Neys, 2012; Toplak et al., 2014b), rather than studies that have not (Morsanyi & Handley, 2008). Children who endorsed more actively open-minded thinking and who performed better on cognitive abilities were also more likely to perform better on belief bias syllogisms (Kokis et al., 2002; Toplak et al., 2014b).

Optimal performance on ratio bias tasks (also termed denominator neglect and attribute substitution) also requires overriding miserly processes. Specifically, miserly processes tend to cue an incorrect response towards selecting the bowl with more marbles (visually appears correct) over the better selection, which is a bowl that has a higher probability of winning (statistically the better option). This task originated from the adult literature (Denes-Raj & Epstein, 1994; Kirkpatrick & Epstein, 1992). Some numerical competence or knowledge is required in order to compare the ratios or probabilities of each bowl. In order to offload the knowledge demands of this task in developmental samples, some studies have explicitly provided the ratios on cue cards so that children do not have to compute the probabilities (Kokis et al., 2002; Toplak et al., 2014b). Some studies suggest that ratio bias performance seems to improve with age (Klaczynski, 2001; Toplak et al., 2014b). Better performance on ratio bias tasks has also been associated with more endorsement of actively open-minded thinking and higher cognitive abilities in developmental samples (Kokis et al., 2002; Toplak et al., 2014b).

In framing problems, participants are asked to make choices in parallel situations where the problem is “framed” differently, such as either positively versus negatively or as gains versus losses. Participants who make different choices based on these re-descriptions are demonstrating a context effect. The classic framing problems used in the heuristics and biases literature involved gain and loss frames (Kahneman & Tversky, 1984, 2000). In developmental studies, children have typically been presented with scenarios involving small prizes that the children receive instead of imaginary deaths. These problems do not seem to require particular mindware, however, consideration needs to be given to understanding how responding to risks and losses unfolds developmentally. This literature in developmental samples has been more complex and inconsistent. Studies have reported framing effects in young children (Schlottmann & Tring, 2005) and in adolescents (Chien, Lin, & Worthley, 1996). Some studies have reported no developmental trends for framing effects (Levin & Hart, 2003; Levin, Weller, Pederson, & Harshman, 2007), but six- to eight-year-old children were found to be more risk averse for gains than for losses in the manner that prospect theory predicts. Reyna and Ellis (1994) reported that four-year-olds displayed no framing effect, eight-year-olds displayed a reverse framing effect, and 11-year-olds displayed a mixture of framing effects. Alternatively, Toplak et al. (2014b) used attribute framing problems in a developmental sample. Older youth were less likely to display framing effects than younger youth and resistance to framing was found to be associated with higher cognitive abilities and endorsement of actively open-minded thinking. Similarly, Weller, Levin, Rose, and Bossard (2012) examined resistance to framing in a preadolescent sample using both gain/loss problems and attribute framing problems. They found that resistance to framing was associated with higher ratings of inhibitory control (a subscale of effortful control) by the preadolescents and their parents.

Temporal discounting and delay of gratification tasks assess a prudent attitude toward the future (Basile & Toplak, 2015; Stanovich, 2009a, 2011; Toplak, Hosseini, & Basile, 2016). These tasks generally involve a choice between the opportunity for a higher gain in the future over a lower immediate gain in the present. Delay of gratification paradigms were studied before temporal discounting in developmental samples. The delay of gratification paradigm was first introduced in developmental samples by Walter Mischel (Mischel, 1973, 2014; Mischel & Ebbesen, 1970), and since that time the research on this paradigm has become well documented and popularized (Mischel, 2014). Temporal discounting paradigms have also been studied extensively in developmental samples (e.g., Steinberg et al., 2009). Temporal discounting paradigms involve multiple trials, unlike the delay of gratification paradigms, and thus have provided the opportunity to better understand performance variability and developmental differences. Temporal discounting tasks require participants to make multiple choices between a small reward immediately versus a larger constant reward available after a delay (Rachlin, Raineri, & Cross, 1991). The size of the immediate reward and length of the delay are varied across trials in temporal discounting tasks. Studies with developmental samples have reported that older youth tend to prefer the larger delayed reward than younger youth (Green, Fry, & Myerson, 1994; Prencipe et al., 2006; Steinberg et al., 2009). The preference for a larger delayed reward has been shown to be positively correlated with performance on intelligence and executive function tasks (Prencipe et al., 2006; Steinberg et al., 2009). Delay of gratification paradigms seem to display a parallel trend; delay ability in preschoolers has been shown to significantly predict academic achievement (SAT scores) in adolescents (Mischel, Shoda, & Peake, 1988).

Another domain of miserly information processing emerges from how our emotions regulate our sensitivity to rewards and punishments. This has been indexed with the well-known Iowa Gambling Task (IGT) that has been studied extensively in adults, as well as in developmental samples. Poor performance on the IGT has been attributed to dysregulation of somatic markers (Damasio, 1994, 1996, 1999). Namely, individuals who perform poorly on this task purportedly have weaker somatic or physiological cues to guide risky choices (Damasio, 1994, 1996, 1999). Somatic markers, or emotions, are suggested to assist by constraining the decision-making space, giving various alternatives preferential availability over other alternatives (Oatley, 1999). While critiques have been written regarding explanations of performance on the IGT and the details of the somatic marker hypothesis (Dunn, Dalgleish, & Lawrence, 2006), the literature on the IGT provided a measurable paradigm to conceptualize how input from emotional modules impact judgment and decision-making (Stanovich et al., 2016). Several developmental studies have demonstrated that older youth tend to make more advantageous selections than children on adapted versions of this task (Crone & van der Molen, 2004; Garon & Moore, 2004; Hongwanishkul, Happaney, Lee, & Zelazo, 2005; Hooper, Luciana, Conklin, & Yarger, 2004; Lamm, Zelazo, & Lewis, 2006; Prencipe et al., 2006). Advantageous selections in this task have also been reported to show modest positive correlations with some executive function tasks (Lamm et al., 2006; Prencipe et al., 2006). However, one study has reported a curvilinear relationship, suggesting an increase in performance between preadolescence and mid-adolescence (10–16-year-olds) and a decrease in performance between mid-adolescence and adulthood (16–30-year-olds; Steinberg, 2010).

There are several other domains of judgment and decision-making tasks indicated in Table 2 that rely heavily on resistance to miserly information processing that have been examined in developmental samples. For instance, myside bias refers to the tendency to test hypotheses in a way that is biased towards one’s own opinions and beliefs (Baron, 1995; Perkins, Farady, & Bushey, 1991; Stanovich, 2011; Toplak & Stanovich, 2003). The myside bias has been shown to be unrelated to cognitive abilities in adults (Stanovich & West, 2007, 2008a). Age effects have not been associated with myside bias in adolescent and young adult samples (Baron, Granato, Spranca, & Teubal, 1993; Klaczynski & Lavallee, 2005; Klaczynski & Narasimham, 1998), but the number of otherside reasons given in a myside bias task has been found to be associated with age in a sample of children and young adolescents (Toplak et al., 2014b). The Cognitive Reflection Test (CRT) has primarily been characterized as a measure indicating miserly information processing (Toplak, West, & Stanovich, 2011, 2014a), but basic numerical competence has also been shown to be an important predictor of task performance (Liberali, Reyna, Furlan, Stein, & Pardo, 2012). Some studies have used CRT items as part of a numeracy scale (Weller et al., 2012), as numeracy skills are required for successful completion of CRT items, but resistance to miserly information processing is additionally required for successful performance on the CRT (Toplak et al., 2011). The CRT has been less well studied in developmental samples, likely attributable to the fact that performance is generally quite low in adult samples (Frederick, 2005; Toplak et al., 2014a). One study has reported that young adults outperformed adolescents on an expanded version of the CRT (Primi, Morsanyi, Chiesi, Donati, & Hamilton, 2015). Overconfidence paradigms have been used to measure performance calibration. For example, better calibration (or less overconfidence) in a sample of preadolescents has been associated with higher ratings of attentional focus and inhibitory control by the preadolescents and their parents (Weller et al., 2012). Older children also tend to provide more accurate estimations of their abilities and competence compared with younger children on other cognitive estimation tasks (Desoete & Roeyers, 2006; Lipko, Dunlosky, & Merriman, 2009; Newman, 1984; Schneider, Visé, Lockl, & Nelson, 2000). Other paradigms that primarily require resistance to miserly information processing have been studied in the adult literature, such as anchoring and outcome bias (Stanovich et al., 2016), but these have been less well studied in developmental samples (Klaczynski, 2001; Smith, 1999).

Reliance on Specific Knowledge or Mindware

Some measures of judgment and decision-making rely heavily on the acquisition of specific mindware or knowledge. The acquisition and consolidation of specific knowledge will make available other potential responses. In order for these responses to be generated or selected, the need for an alternative response must be recognized and adequate cognitive resources must be available to inhibit and sustain cognitive decoupling operations (Stanovich & West, 2008b). Basic practical numeracy skills are critical for many judgment and decision-making tasks, including on some of the tasks that rely most heavily on resisting miserly information processing that have already been discussed. There are individual differences in practical numeracy skills among U.S. children, including basic proficiency with percentages, fractions, and probabilities (Reyna & Brainerd, 2007). Individual differences in numeracy have been found to predict performance on rational thinking and decision-making in adults (Peters, 2012; Peters et al., 2006; Sinayev & Peters, 2015; Weller et al., 2013).

Sensitivity to expected value is another domain of mindware relevant to judgment and decision-making performance (Jasper, Bhattacharya, Levin, Jones, & Bossard, 2013). The “cups” task has been used to examine risky decision-making in developmental samples (Levin et al., 2007; Weller, Levin, & Denburg, 2011). In this task, three variables were manipulated: gain versus loss trials, different levels of probability for the risky choices (0.20, 0.33, 0.50), and different levels of outcomes. Each trial required a choice between a certain or risky option. For example, participants choose between a sure gain of 25 cents or a 20% chance of winning 50 cents. The risky option offered either a higher or lower expected value than the certain option. Feedback was given following each choice, and the accumulated money earned was received at the conclusion of the experiment. It was found that adults were more likely to select the options that offered more favorable expected values than children (Levin et al., 2007; Weller et al., 2011).

Scientific thinking and financial concepts are more specialized types of knowledge that are also important for successful performance on several judgment and decision-making tasks (Stanovich et al., 2016). For example, knowledge about how to test and revise theories, design proper controls for extraneous variables, and objectively evaluate evidence are key concepts of scientific thinking (Zimmerman, 2007). Understanding the need for experimental controls, such as isolating variables and inclusion of control conditions, has been shown to increase with age in children (Klahr, Fay, & Dunbar, 1993; Klahr & Nigam, 2004; Masnick & Morris, 2008; Tschirgi, 1980). Older children are also more likely to take covariation among variables into account and resist inappropriate causal inferences than younger children (Klaczynski, 2001; Koslowski, Condry, Sprague, & Hutt, 1996; Koslowski, Okagaki, Lorenz, & Umbach, 1989; Richardson, 1992). Several developmental studies have examined direct knowledge of these principles. However, if these tasks are designed such that knowledge of the scientific reasoning principle conflicts with an intuitive incorrect response, then the task is better categorized as relying heavily on resisting miserly processing and requiring specific mindware.

Financial literacy involves basic knowledge about saving and investing money (e.g., stocks versus savings accounts), knowledge of economics concepts (e.g., supply and demand), and recognizing sunk costs. Admittedly, issues of knowledge become important in developmental samples, as many children and youth may have a limited understanding of many financial concepts. It may not be developmentally suitable to examine many of these concepts in children. In the limited work that has been done with developmental samples, older participants have demonstrated higher levels of financial literacy and more economic thinking than younger participants (Mandell, 2009; Thompson & Siegler, 2000). Empirical findings on sunk cost effects with developmental samples have been more mixed (Baron et al., 1993; Klaczynski, 2001; Morsanyi & Handley, 2008; Strough, Mehta, McFall, & Schuller, 2008).

Resistance to Miserly Information Processing and Reliance on Specific Knowledge or Mindware

Many judgment and decision-making tasks rely heavily on resisting miserly information processing and specific mindware for successful performance (see Stanovich, 2009a, table 12.1). While cognitive failures on some judgment and decision-making tasks may be largely attributable to miserly information processing or missing mindware, performance on some problems may be attributable to either or both of these difficulties. Attribute substitution problems (Fong, Krantz, & Nisbett, 1986; Stanovich et al., 2008), also referred to as vividness effects, require knowledge of base rates (that large sample information is more diagnostic than single-case testimonies), and the tendency to select the vivid, salient personal testimony must be overridden in order to derive an optimal response. Several studies have reported that increasing age is associated with more reliance on base rates and less reliance on salient vivid cases in older youth than in younger children (Davidson, 1995; Klaczynski, 2001; Kokis et al., 2002; Toplak et al., 2014b). Cognitive ability has also been reported to be significantly positively associated with base rate usage (Kokis et al., 2002; Toplak et al., 2014b). However, some studies have reported increased reliance on salient vivid cases in older participants when social stereotypes are involved (Davidson, 1995; De Neys & Vanderputte, 2011; Jacobs & Potenza, 1991).

Conjunction effects, the gambler’s fallacy, and sample size problems are other examples of these types of problems that require specific mindware and resistance to miserly processing. Conjunction effects refer to the tendency to select options that suggest that the conjunction of two events is statistically more likely than the individual event. The gambler’s fallacy refers to a violation of the principal of independence in probability theory, that a coin is statistically more likely to land on tails after five tosses than recognizing that the probability of each outcome (heads or tails) is statistically equal on all tosses. In sample size problems, outcomes based on larger samples should be recognized as providing more diagnostic information than smaller sample sizes. Older youth have been shown to outperform younger youth on the gambler’s fallacy and sample size problems (Chiesi, Primi, & Morsanyi, 2011). However, the findings for conjunction problems with developmental samples have been mixed (Chiesi et al., 2011; Davidson, 1995; Fishbein & Schnarch, 1997; Klaczynski, 2001; Morsanyi & Handley, 2008). The understanding of the conjunction effect involves sophisticated knowledge for developmental samples. Notably, the conjunction effect using within-subject designs has not been consistently associated with individual differences in cognitive abilities in adult samples (Stanovich & West, 1998). These types of tasks, as assessed in the Probabilistic and Statistical Thinking and Scientific Thinking subtests of the Comprehensive Assessment of Rational Thinking (CART) (Stanovich et al., 2016), have been the most robust predictors of overall judgment and decision-making performance.

When Mindware Interferes With Judgment and Decision-Making Performance

The concept of contaminated mindware was first described by Stanovich (2004), which was later described as the “problem of contaminated mindware” (Stanovich, 2009a; Stanovich et al., 2008). Stanovich introduced the constituents of contaminated mindware within the framework of memetic theory, whereby a meme—the cultural unit analogous to a gene (Dawkins, 1976)—is a selfish replicator that exists for its own propagation across culture, regardless of its validity and in the absence of benefit to the person that holds it. Specifically, contaminated mindware refers to the accumulation of misinformation and unwarranted beliefs, some of which may stop individuals from engaging in reflection and considering alternatives (Stanovich, 2009a). A goal of rational thinking is to base our decisions and actions on what we know to be true of the world and accurate beliefs, called epistemic rationality (Stanovich, 2016; Stanovich et al., 2016). Pseudoscientific and unwarranted beliefs contribute to individuals’ accumulation of contaminated mindware, and if relied upon can hinder one’s critical thought processes and in turn lead to poor judgments and decisions (Evans & Stanovich, 2013). There are many categories to contaminated mindware, but three that have been studied empirically in the judgment and decision-making literature include paranormal, conspiracy, and anti-science beliefs and attitudes (Stanovich et al., 2016).

These domains have been examined relatively less in children than in adults, but the construct of superstitious thinking has been examined in some developmental studies. These studies have indicated relatively mixed findings, with some studies suggesting negative trends or associations with age (Kokis et al., 2002; Toplak et al., 2014b) and other studies suggesting positive associations with age (Eve & Dunn, 1990; Preece & Baxter, 2000). One study indicated no association between age and magical thinking (Bolton, Dearsley, Madronal-Luque, & Baron-Cohen, 2002). This is an important and interesting area for continued study in developmental samples, to determine how acquired mindware emerges to potentially interfere with judgment and decision-making performance.

Paranormal phenomena generally violate the principles of science or current scientific understanding (Broad, 1953; Tobacyk & Milford, 1983). Paranormal beliefs include a belief in Psi (e.g., psychokinesis), witchcraft (e.g., spells and black magic), superstition (e.g., lucky numbers), spiritualism (e.g., communication with the deceased), precognition (e.g., astrology predicting the future), extraordinary life forms (e.g., the abominable snowman of Tibet), and supernatural phenomena (Lindeman & Aarnio, 2007; Lindeman & Svedholm, 2012). Among paranormal beliefs, superstitious thinking and beliefs have been most well studied in developmental samples. Some empirical studies have reported a negative association between judgment and decision-making performance and endorsement of superstitious beliefs. Specifically, in a sample of children in grades five, six, and eight, endorsement of superstitious thinking was negatively correlated with deductive reasoning, cognitive ability, and actively open-minded thinking (Kokis et al., 2002). In a sample of children in grades two through nine, endorsement of superstitious thinking was negatively associated with cognitive ability, actively open-minded thinking, and performance on several other judgment and decision-making measures (Toplak et al., 2014b). Similarly, in a sample of children in grades two through eight, superstitious thinking was negatively correlated with cognitive ability, need for cognition, and probabilistic reasoning (Chiesi et al., 2011). Further, in samples of adolescents, superstitious beliefs correlated positively with an intuitive style of thinking and negatively with a rational style of thinking (Marks, Hine, Blore, & Phillips, 2008) and was predictive of gambling behavior (Donati, Chiesi, & Primi, 2013).

Paranormal phenomena have been more extensively studied in childhood and adolescence relative to other domains of contaminated mindware (e.g., Eder, Turic, Milasowszky, Van Adzin, & Hergovich, 2011; Hergovich, Schott, & Arendasy, 2008). One potential explanation for the disproportionate attention given to paranormal beliefs early in development as compared with other domains of contaminated mindware is that paranormal beliefs are observed early in development and manifest in magical and supernatural thought and superstitious rituals (e.g., Bolton et al., 2002; Kokis et al., 2002; Preece & Baxter, 2000). On the other hand, conspiracy and anti-science beliefs become relevant by adolescence, as individuals gain better understanding of science and societal concerns, such as conspiratorial thinking.

Conspiracy theories attribute the cause of an event to secret plots by powerful groups or forces (Douglas & Sutton, 2008; Goertzel, 1994; McCauley & Jacques, 1979). These theories are resistant to falsification and lack evidence of their validity (Sutton & Douglas, 2014), despite sometimes being true. A general conspiracist belief reflects “the unnecessary assumption of conspiracy when other explanations are more probable” (Aaronovitch, 2009, p. 5). Conspiracist belief rejects and deflects criticism, labeling any criticism as part of the conspiracy, further confirming the conspiracist belief system (Boudry & Braeckman, 2012). There is evidence for the prominence of conspiracy beliefs as early as adolescence (e.g., Bogart & Thorburn, 2006; Thorburn & Bogart, 2005). Thorburn and Bogart (2005) discuss the role of sociocultural and historical factors in creating a distrust in the government and medical institutions, which precedes and reinforces conspiracy beliefs as early as adolescence. Anti-science beliefs refer to the rejection of and opposition to the scientific method (Holton, 1992, 1993). General anti-science attitudes and beliefs include seeing science as possessing low credibility (Hartman, Dieckmann, Sprenger, Stastny, & DeMarree, 2017), with a preference for intuition and instinct (Stanovich et al., 2016). Similar rates of endorsement of conspiracy and anti-science beliefs across adolescence and young adults have been demonstrated (Rizeq, Flora, & Toplak, in press).