In the following, learning with worked examples and the relation to cognitive load is described. Afterwards different kinds of worked examples are introduced and their usefulness for fostering diagnostic competence is analyzed.
Worked examples are composed of a problem formulation, solution steps (which may be more of less detailed), and a final solution. There is evidence for the effectiveness of worked examples in well-structured domains (e.g. in mathematics) (Renkl &, 2010; Stark, 1999, 2001) and also in complex domains (e.g. in argumentation or legal case reasoning) (Nievelstein, van Gog, van Dijck, & Boshuizen, 2013; Schworm & Renkl, 2007). The worked example effect can be explained with the cognitive load theory (Kalyuga, 2011; Sweller, Van Merriënboer & Paas, 1998). For the elaboration of a worked example, less cognitive capacity of the working memory is demanded than problem solving. This effect can be assumed to last until the learner has gained sufficient expertise and therefore acquired enough cognitive schemata to lead his or her problem solving processes (van Merriënboer, 2013). As a result, more cognitive capacity is available for the construction of schemata; that is, to build meaningful relations between prior knowledge and new information (Kalyuga, 2011). If learners already have much prior knowledge, an expertise reversal effect can occur while learning from worked examples (Kalyuga, Ayres, Chandler, & Sweller, 2003). An expertise reversal effect occurs if an instructional method is no longer beneficial for learners with a higher level of knowledge (Kalyuga & Renkl, 2010). Before learners can start to benefit more from problem solving than from learning with worked examples they need to acquire knowledge on the domain principles (Renkl, in
Chapter 3: Instructional Support for the Acquisition Diagnostic Competence
press). In complex domains there are studies in which no expertise reversal effect was found for learning with worked examples: for instance, with legal case reasoning (Nievelstein et al., 2013), whereas in other complex domains such as literacy interpretation an expertise reversal effect was found (Oksa, Kalyuga, & Chandler, 2010).
Different types of worked examples can be distinguished: (1) product-oriented worked examples, (2) process-oriented worked examples and heuristic worked examples and (3) double-content worked examples. (1) In classic product-oriented worked examples it is shown how a goal-state can be reached. (2) In process-oriented worked examples, the strategic knowledge on heuristics and problem-solving strategies applied to reach the goal is included in addition; also the rationale of a solution is also explained (van Gog et al., 2004). Process-oriented worked examples are promising in order to increase transfer (van Gog, Paas, & van Merriënboer, 2008; van Gog et al., 2004). A similar concept is that of heuristic worked examples (Hilbert, Renkl, Kessler, & Reiss, 2008). Similar to process- oriented worked examples, problem-solving strategies are added but only for non-recurrent skills (ibid). Therefore they can be regarded as a special type of process-oriented worked examples. Heuristic worked examples were effective in learning to prove (ibid). (3) Another form of worked examples to foster complex skill are double-content worked examples (Schworm & Renkl, 2007). To foster the development of e.g. argumentation skills it is required to have two levels of content that is argumentation itself (learning domain) and also the domain from which the problem is taken e.g. genetics (exemplifying domain) (Schworm & Renkl, 2007). Further, in argumentation no algorithmic solution can be provided. Double-content examples were successfully used to foster argumentation (Schworm & Renkl, 2007) and collaboration (Rummel, Spada, & Hauser, 2009; Rummel & Spada, 2005).
For the diagnosis of patients and also for the diagnosis of a classroom situation it seems that there is no algorithmic solution available, as these kinds of problems can be regarded as highly complex and ill-defined (see chapter 2.1.4 Differences and Similarities of Diagnostic Competence in the Domains). Accordingly, using the principles from process-oriented examples might be beneficial to foster diagnostic competence. Thus, knowledge on heuristics and problem-solving strategies (strategic knowledge) and also of the rationale of a solution (conditional knowledge) should be included. Similar to the double-content examples for argumentation or collaboration, in diagnosing there is also an exemplifying domain where basic features have to be understood, e.g., while diagnosing a patient with symptoms of cardiac failure basic declarative-conceptual knowledge on the cardiovascular system needs to be understood. However, it seems that the distinction between the exemplifying and content domains is less clear and much more interwoven.
An assumption of worked examples is the principle that learners can also learn by observing others problem solving (Sweller, 2010). That individuals can also learn from
Chapter 3: Instructional Support for the Acquisition Diagnostic Competence
others’ cases is also assumed in the case-based reasoning approach (Kolodner, 2006) and in the social learning theory (Bandura, 1977). Learning from others was so far investigated from a cognitive perspective in research on worked examples and from a social learning perspective in research on modeling examples (van Gog & Rummel, 2010). Van Gog and Rummel (2010) contrast worked examples to modeling examples by describing the solution to a problem of worked example as didactically motivated. In modeling examples the model can also be a peer who shows natural behavior and commits errors while solving a problem. Worked examples are generally presented in a text-based format whereas modeling examples often uses some kind of live or captured observations (van Gog & Rummel, 2010). In both perspectives it is assumed that learners need to actively process the example cases and build on cognitive representations (ibid). Learning from worked examples and learning from observation share common features, such as reliance on cases (Renkl, in press). Both strive to build activities that let the learner build relations between the cases and the underlying principles. Renkl (in press) states that in several studies worked examples and observational learning are conceptually connected (Chi, Roy, & Hausmann, 2008; Craig, Chi, & VanLehn, 2009; Gholson & Craig, 2006). In these studies, a learner observes another learner trying to perform a skill while the observed learner is tutored. As the learner that is observing is not addressed personally it can be regarded as a case of vicarious learning (McKendree, Stenning, Mayes, Lee, & Cox, 1998).
It might also be beneficial to connect principles from worked examples and observational learning to foster diagnostic competence. Using a fictitious peer in a worked example format could have the advantage of increased transfer performance, as the situation of a peer doing an internship at a school or a medical clerkship in a hospital is much closer to a situation a student will encounter soon. Another potential advantage would be that misconceptions and typical errors could be integrated more authentically. Through including a fictitious expert, strategic and conditional information can be added into the worked examples.
To achieve meaningful learning in the form that new knowledge is integrated into existing knowledge structures, active processing of learning material is necessary (Eysink & de Jong, 2012). Accordingly, the effectiveness of worked examples is dependent upon the self-explanation activity of a learner (Atkinson, Renkl, & Merrill, 2003; Chi & Bassok, 1989; Hausmann & VanLehn, 2007; Renkl, 1997). Self-explanation means to generate explanations after being confronted with learning material (Chi, 2000). With regard to worked examples, that is, if a learner can and does explain the solution steps to him or herself. Self-explanations are not complete but rather fragmented, incorrect, and incomplete and thus show what a learner did and did not understand (Chi, 2000). Incorrect self-explanation can also promote learning if detected and resolved (Chi, 2000). Conati and VanLehn (2000) in contrast state that only correct and high quality self-explanation are
Chapter 3: Instructional Support for the Acquisition Diagnostic Competence
learning process. Accordingly, if an incorrect self-explanation cannot be detected because a learner has insufficient monitoring skills, it might not be advantageous for learning.
Self-explanations are necessary to gain understanding (Nokes, Hausmann, VanLehn, & Gershman, 2011), and can improve transfer (Atkinson et al., 2003; Hilbert et al., 2008). Differences exist in the success of learners due to qualitative differences in their self- explanation activity (Chi, Bassok, Lewis, & Reimann, 1989; Renkl, 1997); for example successful learners anticipate the next solution step and connect underlying principles within the case (Renkl, 1997).
Without support, the cognitive capacity freed through instruction based on worked examples is not used for self-explanation by all learners (Renkl, 1997; Stark, 1999). Instead, learners often process worked examples passively or superficially (Renkl & Atkinson, 2010). Self-explanation activities can be fostered indirectly through the design of the worked examples (Renkl, in press). A promising method to help learners use this capacity for learning is include errors into worked examples (Booth et al., 2013; Große & Renkl, 2004, 2007; Stark et al., 2011). For worked examples that are conceptually connected to observational learning by using a peer as fictitious model, this can easily be realized.