Dual-process Models1
A Dual-Process Model of Adolescent Development: Implications for Decision Making, Reasoning, and Identity
Paul A. Klaczynski
The Pennsylvania State University
(in press). In R. V. Kail (Ed.), Advances in Child Development and Behavior, Vol. 31. San Diego, CA: Academic Press.
I. Introduction
II. Dual-process Theories of Cognition
III. Developmental Evidence for Two Processing Systems
IV. The Development of Conditional Reasoning
V. The Development of Decision Making Heuristics and Decision Making Competencies
VI. Development and the Belief-motivation-reasoning Interface
VII. Identity Formation, Belief-biased Reasoning, and Metacognitive Dispositions
VIII. Conclusions: What Develops?
References
I. Introduction
The last 20 years of research on adolescent development in the 20th century has witnessed two disconcerting trends. First, research on cognitive development and research on social development have become increasingly independent. One result of this increased independence has been theories of social development that attribute psychological adjustment and age-related changes in social development primarily to forces in adolescents’ environments (e.g., family, peers) and that give lip service, at best, to the role of cognition. Second, inspection of programs from prestigious conferences, such as those held by the Society for Research in Child Development and the Society for Research on Adolescence, indicates that, if research on adolescent cognitive development is being conducted, it is being presented far less frequently than research on social development. Adolescent textbooks reflect both of these trends. On the one hand, the sole cognitive theory presented in most texts that is a truly developmental is Piaget’s theory of formal operations. On the other hand, chapters on adolescent social behavior do not typically go far in integrating recent theories of and findings concerning adolescent cognition.
In this present chapter, I present evidence for a dual-process theory of cognitive and social-cognitive development. The theory is potentially applicable to all age groups, but most of the relevant research has involved comparisons between children and adolescents or has focused solely on adolescents. Although the adolescent focus of dual-process research necessarily limits the extent to which generalizations about the mechanisms underlying cognitive development can be made to other age groups, this focus also provides a foundation for the construction of bridges between adolescent cognitive and social development. In part, then, I hope to build a bridge that connects aspects of social, cognitive, and social-cognitive development that currently exist as isolated islands. As a prelude to my discussion, I believe that the model advanced here has implications for understanding decision making, in-group and out-group biases, stereotyping, reasoning, and the self. The more specific goals of this chapter are to discuss research on “belief-biased reasoning,” decision making, conditional reasoning, and identity formation. The current running through and tying together these discussions is the belief that dual-process theories can provide useful accounts of developments in each area.
The chapter is divided into the following sections: First, in Section II, I present a brief overview of a dual-process theory of cognitive and social-cognitive development. This theory, which has much in common with similar theories in social and cognitive psychology, operates under the assumption that cognition is guided by processing in two independent systems. Following this introduction to dual-process theories, in Section III I briefly review of some of the research that supports the value of adopting a dual-process approach. This section is followed by a discussion, in Section IV, of recent research on the development of conditional reasoning. In Section V, I present results from a series of investigations into children’s and adolescents’ decision making. In Section VI, I focus on a model of belief-biased reasoning, an outline of relevant research, and the relationship between biased reasoning and “thinking dispositions.” In Section VII, I outline new research exploring the relationships among belief-biased reasoning, epistemological development, and adolescent identity formation. I conclude, in Section VIII, by discussing the question how adolescent cognition differs from that of children and of how dual-process theories can be usefully employed in be in framing future research on adolescent social and cognitive development.
II. Dual-process Theories of Cognition
In literature on the adult reasoning and decision making, perhaps the most striking finding of research conducted between 1972 and 2004 is that, on a variety of tasks and under a wide range of conditions, performance falls well short of traditional prescriptions for normative responding (Kahneman & Tversky, 1996; Piatelli-Palmarini, 1994). The large discrepancy between how people “should” respond and how they actually respond has been referred to as the “normative/descriptive gap” (see Baron, 1988; Stanovich & West, 2000). Although the interpretation that these findings are indicative of basic shortcomings in adult reasoning and decision making competence has been challenged on a number of grounds (e.g., Cohen, 1981; Gigerenzer, 1996; Hilton, 1995), the list of purported “biases” in human reasoning is impressive. Research in the so-called “heuristics and biases” literature shows that adults frequently commit the “conjunction fallacy,” ignore base rate information, make unrealistically optimistic judgments, overemphasize the usefulness of vivid, episodic memories in making judgments, are unduly biased by prior beliefs, make systematic errors on simple ratio problems, and are prone to numerous other errors (for reviews, see Evans & Over, 1996; Kahneman, Slovic, & Tversky, 1982; Stanovich, 1999).
To illustrate the normative descriptive gap, consider the following “conjunction” problem (a variant of the original “Linda” problem used by Tversky & Kahneman, 1983):
Linda is 31 years old, single, wears glasses, and wears dresses that are out of fashion. Although her hair is somewhat long, she usually keeps it in a tight bun. Linda enjoys listening to music and reading books. As a college student, she was deeply concerned with social issues, such as discrimination, poverty, and social justice. She also participated in a number of political demonstrations.
After reading this description, participants rank the following statements in order of the likelihood that they accurately describe Linda.
a. Linda is a teacher in an elementary school
b. Linda is a bank teller
c. Linda works in a bookstore and takes Yoga classes
d. Linda is active in the feminist movement
e. Linda is a member of the League of Women Voters
f. Linda is a bank teller and is active in the feminist movement
The “conjunction fallacy” occurs when the probability that Linda is both a bank teller and a feminist is ranked higher than the probability that Linda is a bank teller. In the “strong” version of the fallacy (see Klaczynski, 2001a), the conjunctive description (f) is ranked higher than both of its components (b and d). However, in any conjunction, p(AB: bank teller and feminist) cannot exceed p(A: bank teller) or p(B: feminist) because the individual categories (A and B) necessarily include all subcategories (e.g., bank tellers who are feminists must be members of the superordinate category, "bank tellers"). Although the conjunction task involves little more than class inclusion, most adults indicate that option “f” is more likely than option “b” and a significant minority believe that “f” is more likely than both “b” and “d.”
Debates over the normative response to this and other problems (e.g., the normative response to Wason’s [1966] selection task has been subjected to similar scrutiny) have been fierce (for review, see Stanovich, 1999) because, if responses traditionally judged as normative are accepted as such, then it would appear that most college-educated adults are poor decision makers and reasoners and, therefore, are fundamentally irrational (see Piatelli-Palmarini, 1994). Among the most prominent arguments supporting the case for “fundamental irrationality” is that, because humans have limited information processing capacity (and, in particular, because the size of working memory is limited), they must “satisfice” (Simon, 1993) in their decision making. That is, because information processing capacities are easily overwhelmed by problem complexity, people do the best they can with their limited resources.
As appealing as this “bounded rationality” argument is to many cognitive scientists, it has been criticized on the grounds that it does not explain why people sometimes make decisions that are in almost direct opposition to the decisions that are normatively prescribed and it cannot account for some of the developmental evidence described subsequently. Bounded rationality arguments also cannot explain why people often make errors on simple ratio problems that impose few demands on cognitive resources (Reyna, 2000). Furthermore, these arguments seem to ignore the distinction between competence and performance. People may well possess the capacity to make normative judgments and decisions on a variety of tasks in the heuristics and biases literature, but may not often evince that competence in performance.
Dual-process theories of cognition have arisen as alternatives to traditional information processing theories to account for these perplexing findings—that is, dual-process theories hope to explain why college students perform poorly on decision making and reasoning tasks that, at least on the surface, should be solved easily. Specifically, surface characteristics of problems may not necessarily overwhelm information processing capacities, as assumed by bounded rationality theorists. Instead, task characteristics may determine which of two independently functioning information processing systems is predominant on that task. One system—here called the “analytic” system—has been the traditional focus of both cognitive psychologists and cognitive developmentalists. When it is predominant, college students’ solutions to many heuristics and biases tasks are often normative. However, under most circumstances a different system—here labeled the “experiential” system—is predominant. Because this system relies more heavily on procedural and episodic memory than on active computation of choices, alternatives, and reasons, solutions produced under the predominance of this system are sometimes at odds with normative prescriptions.1
The experiential system involves the preconscious activation of procedural and episodic memories that can be used to guide judgments and decisions (Chen & Chaiken, 1999; Epstein, 1994; Epstein & Pacini, 1999). Consider the “Linda” problem. Most people believe that Linda in a bank teller and a feminist because problem content cues stereotypes of both categories. Rather than relying on logical processing, computation of probabilities, or construction of Venn diagrams, people instead base their judgments on strongly activated memories. On problems such as this, people are believed to rely on heuristics short-cuts in making their judgments. Thus, the “representativeness” heuristic—the belief that a target is prototypical of a particular category —is activated when the information about Linda is presented (see Kahneman, Slovic, & Tversky, 1982 for a discussion of other tasks that involve the representativeness heuristic).
In general, experiential processing is fast and operates automatically, at the “periphery” of consciousness (Epstein, 1994). This system facilitates information mapping onto and assimilation into existing knowledge categories, operates to convert conscious strategies and tactics into automatic procedures and strategies and aids the activation of decision-making heuristics and other memories (e.g., beliefs, vivid episodic memories) that bias judgments and interfere with attempts to reason “objectively.” Because it likely evolved before the analytic processing system and, more importantly, because it requires little cognitive effort and expends few cognitive resources, experiential processing is often considered the overall system’s default (Brainerd & Reyna, 2001; Epstein, 1994; Stanovich, 1999).2
Thus, experiential processing depends on the activation of heuristic short-cuts, most of which are acquired through experience. Developmentally, this means that individuals’ repertoire of heuristics should become increasingly diverse and increasingly easily activated with age. The implication of this conclusion is not that adults will use heuristics more than children, but instead that—when experiential processing is predominant—adults’ judgments and decisions will reflect more variability in the types heuristics they use. If children have not yet acquired the heuristics adults typically use on a task, the (possibly mistaken) conclusion that adults rely on experiential processing more than children may be drawn. However, simply because adults have more heuristics available than children does not mean that they will use these heuristics more often. Indeed, knowing that an increasingly diverse repertoire of heuristics is acquired from childhood through adolescence and adulthood suffices to explain neither the frequency with which heuristics are applied to judgment and decision situations nor occasions on which heuristics, although activated, are not exercised. As discussed subsequently, because a particular heuristic is stored in procedural memory does not mean that it will be used when it is activated. The experiential processing system, functioning with little or no conscious awareness on the decision maker’s part, continuously assimilates information and matches internal and external cues to memory procedures; this matching process, in turn, activates and makes available specific heuristics for utilization.
However, the experiential processing system co-develops with the analytic processing system. Indeed, if not for the co-development of the more deliberate analytic system, judgments might well be dominated by “off-the-cuff,” automatically activated and employed heuristics and biases. The analytic processing system comprises consciously-controlled, effortful thinking, and the numerous competencies that have traditionally been considered essential to cognitive development and normative decision making (Evans & Over, 1996; Stanovich, 1999). Unlike experiential processing, analytic processing is directed toward breaking down problems into their component elements, examining these elements, and, from this analysis, deriving a problem solution, judgment, decision, or argument. In further contrast to experiential processing, analytic processing operates on "decontextualized" representations. The process of decontextualization, in turn, is essential if other analytic competencies are to be engaged consistently and effectively (Stanovich, 1999; Stanovich & West, 1997). Decontextualized task representations—wherein the underlying structure (e.g., logical components) of a problem is decoupled from superficial contents (e.g., counterfactual information)—provide a working memory structure on which logico-computational processing can operate (Stanovich & West, 1997; see also Donaldson, 1978). However, the ability to consciously decontextualize task structure and requirements from superficial task contents and misleading memories depends largely on the development of the metacognitive and executive function abilities (e.g., planning, impulse control, ability to inhibit memory-based interference). In Table 1 (adapted from Epstein, 1994; Evans, 2002; and Stanovich, 1999), a brief list of the attributes of the two processing systems is presented.
Because the instantiation of analytic competencies in performance is often highly effortful, if they are to benefit developing individuals, their acquisition must be accompanied by increases in the tendency to consciously employ them. As recent discussions of metacognitive development (e.g., Kuhn, 2000; Moshman, 1990, 1999) and “thinking dispositions” (e.g., Stanovich & West, 2000) highlight, for everyday reasoning and decision making to approach normative ideals, development must proceed beyond the abilities to inhibit memory-based interference, reflect on the processes of reasoning and decision making, and evaluate the quality of decision options. Specifically, developments in analytic competence must be coupled with the acquisition of the dispositions (i.e., personal qualities, such as the “Need for Cognition” [the tendency to seek and enjoy intellectual challenges; see Cacioppo, Petty, Feinstein, & Javis, 1996]) that increase individuals’ inclinations to use these abilities.
As noted previously, development is in part characterized by the acquisition of judgment and decision heuristics. Although some of these heuristics may be learned explicitly, by and large they are acquired through implicit cognitive processes (see Reber, 1992). Once acquired, judgment and decision heuristics are activated automatically by situational cues. Many people also employ these heuristics automatically not only because they are “fast and frugal” (Gigerenzer, 1996), but also because they often lead to outcomes beneficial, or at least not harmful, to the decision-maker. Also, because people have only a fleeting awareness that they have been activated, and because their activation elicits intuitions or “gut” feelings that they are “right” for the immediate situation, decision heuristics are often used in situations for which their relevance is dubitable. Yet, although heuristic activation is effortless and automatic, once activated, it is likely that some (but perhaps not all) heuristics are least momentarily available in working memory. This availability affords reasoners the opportunity to consciously reflect on the value of the heuristic and actively decide whether to use the heuristic or not. As the adult literature indicates, either most people do not engage in this type of conscious reflection or, if they do, most people decide that the heuristic is in fact worth using.
The first point of this discussion is that experiential processing tends to predominate people’s thinking. Second, experiential processing predominance can be overridden by analytic processing (Stanovich, 1999). The process of overriding experiential processing is conscious, likely requires advanced executive function and metacognitive abilities, and therefore is likely to be achieved more effectively by adolescents than by children. However, the third point of the foregoing discussion is that most adolescents (and most adults) are not predisposed to override the experiential system functioning; that is, there appear to be wide individual differences in the tendency to inhibit the utilization of automatically activated heuristics, engage in logical analysis, and construct decontextualized task representations (Stanovich & West, 2000).
This final point leads to an important distinction between metacognitive abilities and metacognitive dispositions. The former comprises a cluster of competencies that involve the ability to effectively reflect on what one knows, how one knows, and how accurate one’s knowledge is; the ability to track the course of one’s reasoning and monitor it for consistency and quality; and the ability to reflect on one’s arsenal of decision and reasoning strategies to determine the most optimal strategy for a given situation. Metacognitive dispositions, by contrast, are motivational in nature and generally reflect beliefs about the value of engaging in effortful, logical analysis and of being aware of the reasoning process. This distinction recalls again the competence/performance distinction: Two individuals may possess similar degrees of metacognitive competence and may therefore be similarly able to inhibit experiential system predominance. But one of these individuals may be more disposed to expending the effort required to achieve the precision enabled by this intercession. The role of metacognitive intercession in regulating analytic processing and in interactions between the analytic and experiential systems is illustrated in Figure 1.