Decreasing Cognitive Load for Novice Students: Effects of Explanatory versus Corrective Feedback in Discovery-Based Multimedia (Moreno, 2004)

Moreno, R. (2004). Decreasing Cognitive Load for Novice Students: Effects of Explanatory versus Corrective Feedback in Discovery-Based Multimedia. Instructional Science, 32(1), 99-113.

This paper investigated the guided feedback hypothesis, which states that in discovery learning environments, explanatory feedback (EF) is superior to corrective feedback (CF) in guiding novice learners. Two experiments, conducted with a botany multimedia learning game, using guidance as scientific explanation, and college students as test subjects, confirmed the hypothesis. Results of two experiments demonstrated both a learning effect and a cognitive load reduction effect due to guidance in the form of EF.[Note: Much of this material is copied directly from Moreno’s paper]

Guided feedback hypothesis – discovery learning environments that use explanatory feedback (EF) to guide novice students in the process of meaning making promote deeper learning than those that present identical materials using corrective feedback (CF) alone. EF reduces the extraneous cognitive load that results from having novice students search for a plausible explanation to the correctness or incorrectness of their problem solutions in discovery environments. This, in turn, frees capacity for schema acquisition and automation, which eventually promotes learning.

“Despite the alleged advantages to using discovery-based multimedia environments to help students in their process of meaning making, cognitive load theory learning suggests that the free exploration of a highly complex environment may generate a heavy cognitive load that is detrimental to learning (Sweller, 1999; Paas, Renkl, & Sweller, 2003). This is particularly important in the case of novice learners, who lack proper schemas to integrate the new information with their prior knowledge (Tuovinen & Sweller, 1999). Thus, whereas on the one hand, cognitive theory encourages the use of discovery-based multimedia programs to promote deep understanding by actively involving students in the learning task, on the other hand, the limited capacity assumptions of cognitive load theory discourage the use of multimedia programs that overload the novice learner with too much new information and no guidance” (p. 100). The main goal of this paper was to examine the role that software pedagogical agents’ feedback may have in solving this conflict.

When students learn science in classrooms with pure-discovery methods and minimal feedback, they often become lost, frustrated, and their confusion can lead to misconceptions (Brown & Campione, 1994; Hardiman, Pollatsek, & Weil, 1986). In a study on one-to-one tutoring interactions, Chi and colleagues (Chi, Siler, Jeong, Yamauchi, & Hausmann 2001) found that when tutors provided prompts alone, students learned as effectively as when tutors gave them explanations and feedback, a finding that was interpreted as a result of the greater effort invested by students to take control of their own learning. [Comparing feedback types with other types of scaffolding in terms of achieving desirable internal learning states in the user may be an important future line of work…].

According to a cognitive theory of multimedia learning (Mayer & Moreno, 2003), a successful learning process includes students’: (p. 101)

  1. Meaningful interaction with academic materials (Moreno, Mayer, Spires, & Lester, 2001);
  2. Selection of relevant verbal and non-verbal information (Paivio, 1986);
  3. Organization of information into corresponding mental models or representations (Mayer & Moreno, 2002; Moreno & Mayer, 2002); and
  4. Integration of new representations with existing knowledge (Pressley, Wood, Woloshyn, Martin, King, & Menke, 1992).

Guided discovery vs. pure discovery environments: In guided discovery environments, a pedagogical agent may facilitate students’ selection, organization, and integration of materials by providing explanatory feedback on their choices. Conversely, in pure discovery environments this role may be absent or limited to providing students with minimum feedback (i.e., if a response is correct or incorrect). Students in pure discovery environments need to select, organize, and integrate the multiple pieces of information on their own. According to cognitive load theory, when students lack significant prior knowledge, the demands that arise from processing the new information without guidance can be overwhelming and leave students with insufficient capacity for
building a coherent mental representation of the system to be learned (Sweller,
1999).

Moreno conducted two experiments testing the guided feedback hypothesis, using a botany multimedia learning program. The effects of EF and CF were compared on retention, problem-solving transfer, program rating and ease of learning condition.

Compared to students who learn with CF alone, the guided feedback hypothesis predicts (p. 103):

  1. Students who learn with EF will remember and transfer better what they have learned to new problem solving situations.
  2. Students who learn with EF will report lower levels of difficulty and higher levels of helpfulness to understand the instructional materials than those who are not guided with EF.
  3. EF treatment will be more effective than the CF treatment, as a consequence of the reduced cognitive load and increased learning performance.

Experimental multimedia program:

The EF and CF versions were based on a multimedia program called “Design-a-Plant” (Lester, Stone, & Stelling, 1999). Content concerned how to design plant parts (roots, stems, and leaves) to survive in different weather conditions (rain, sun, wind, etc.). Pedagogic agent offered spoken explanations concerning the relation between plant parts and weather conditions by providing students with feedback on the choices they make in the process of designing plants. For a set of five different environmental conditions, in the EF version of the program the agent had the following functions: (a) asked the student to select the plant part that was appropriate for each environment from a presented menu, (b) gave feedback on students’ correctness of choice, (c) explained why the chosen plant part was or was not correct such as: “Yes, in a low sunlight environment, a large leaf has more room to make food by photosynthesis” or “Mmm, your deep roots will not help your plant collect the scarce rain that is on the surface of the soil”, (d) showed students the right choice and moved on to the next step. The CF version had identical functions to the EF version with the exception that it did not include step (c).

Cognitive load measurement: “Because self-ratings on the levels of difficulty (mental effort) of the computer game are assumed to assess cognitive load indirectly (Paas, Tuovinen, Tabbers, & Van Gerven, 2003), Paas and Van Merriënboer’s (1993) method to relate mental effort to performance measures was used to calculate the instructional effectiveness of each treatment. This method allows measures of cognitive load to be combined with measures of performance to derive information on the relative effectiveness of the instructional conditions used in the study. Therefore, we used participants’ mean ratings of difficulty in combination with the mean retention and transfer tests for each treatment, to calculate their respective effectiveness” (p. 106).

Results of experiment 1 support the guided feedback hypothesis for novice learners. Low-prior knowledge students achieve better retention and transfer scores, and rate the difficulty and helpfulness of the instructional program more
favorably when the agent provides explanatory feedback during the process
of knowledge construction rather than when the agent provides corrective
feedback alone.

Experiment 2 replicated the conditions of experiment 1 but included data on the number of correct answers each student gave. Result: Group EF gave significantly fewer wrong answers than Group CF while problem solving with the multimedia game. Based on a cognitive theory of learning, students in group CF would give more wrong answers during their interactions because, in the absence of necessary guidance, novice learners resort to problem-solving search strategies that are cognitively inefficient (i.e., trial-and-error) due to a heavy working memory load (Sweller, 1999).

“Interestingly, across both experiments, students who learned with greater or lesser degrees of discovery (CF or EF, respectively) did not report different levels of motivation or interest. That is, consistent with the predictions of a cognitive theory of multimedia learning and cognitive load theory, the results of Experiments 1 and 2 suggest that the way that explanatory feedback benefits students’ learning resides in the cognitive and not the affective variables that are associated with discovery-based multimedia programs” (p. 109). [However, cognitive-affective factors may still be at play…]

These results parallel those of the CLT literature on worked examples,
which show that when students’ preexisting knowledge level is limited,
the presentation of worked examples reduces the extraneous cognitive load
imposed by means-ends problem solving (Sweller et al. 1998). (See Tuovinen & Sweller, 1999).

“On the practical side, the results have direct implications for the design of agent-based learning environments, which is still mostly based on the intuitive beliefs of designers rather than on empirical evidence” (p. 110).

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: