**Introduction:**

My third principle to practice blog post focuses on a paper by:

Salden, Aleven, Schwonk and Renkl (2009)

THE EXPERTISE REVERSAL EFFECT AND WORKED EXAMPLES IN TUTORED PROBLEM SOLVING

**What was the paper about?**

Just a couple of definitions before we start:

*Expertise reversal effect:* as learners become more knowledgeable in a domain, guided instruction becomes less impactful and may actually hinder learning (see the work of Slava Kalyuga for more information).

*Procedural knowledge:* knowing about processes and procedures (e.g. how you might calculate ½ + ¾ = 1 ^{1}/_{4}).

*Conceptual knowledge:* knowing why certain processes work; understanding the relationship between things (e.g. I also know that ½ = 0.5 and ¾ = 0.75, therefore 1.25 = 1 ^{1}/_{4}).

There is a wealth of research to support the use of worked examples to support problem solving, particularly with novice learners because it helps to reduce the overload on working memory. The expertise reversal effect is a notion that suggests that worked examples are more favourable in earlier stages of learning, while problem solving could be more effective in later stages. The authors suggest that a key implication of expertise reversal for instructional design is that worked-out steps should gradually be ‘faded’ from worked examples, rather than jumping from fully worked examples to independent problem solving. They inform us that worked examples need to be adapted to the learner’s level of expertise and only when there is sufficient understanding of key principles should they progress to problem solving, but determining the transition point is a challenge,

**What was the aim of the paper?**

The authors ask the question, when is it more effective to provide assistance (e.g., example solutions), and when is it more effective to let the learner try to generate or construct solutions for themselves (or with lower levels of assistance)?

They address this question by investigating the effectiveness of different worked example fading within an instructional approach using cognitive tutor software. It was hypothesized that an adaptive fading process would lead to better learning than independent problem solving and fixed fading worked examples (more on the difference between these below).

**What did they do?**

Two experiments were conducted (lab and classroom based) using a Cognitive Tutor (computer software) where three conditions were compared: Problem Solving, Fixed Fading Worked Examples and Adaptive Fading Worked Examples. Problems were sequenced from simple to more complex, with one-step problems presented first (procedural), followed by two-step problems, and eventually by three-step problems (conceptual).

This experiment was conducted twice. Once in a lab (n=57; 14-16 years old) and once in a vocational school classroom (n=20; 14-15 years old). The students were randomly assigned to one of three conditions:

In the *Problem Solving* condition, all steps of all problems were ‘pure problem solving’ with learners working independently to progress from procedural to conceptual questions.

In the *Fixed Fading* condition learners started out with fully worked examples, with example steps gradually being faded in subsequent problems until, in the last two problems, all steps were pure problem solving (conceptual).

In the *Adaptive Fading *condition, the presentation of worked steps was the same as in the fixed fading condition up until more challenging problems. Once students reached those problems, any step could be presented as either pure problem solving or worked-out step, depending on the student’s performance in explaining worked-out steps in earlier problems.

Immediate and delayed test (1 and 3 weeks respectively) were administered.

** ****Experiment 1 – Lab based**

Results indicate that *adaptive fading was more effective than the other conditions* on both immediate and delayed post-tests. Additionally, the adaptive fading condition *required fewer worked *steps than the fixed fading condition, indicating that students’ knowledge levels increased faster in the adaptive condition.

**Experiment 2 – Vocational classroom**

Results indicate that the *adaptive fading was more effective than the other conditions* on the delayed post-test. Additionally, the adaptive fading condition required *more examples* than the fixed fading condition, which possibly indicates that overall the students’ knowledge levels increased slower in the adaptive condition.

**What is the key principle of the paper?**

Adaptive faded worked examples are likely to improve novice learners’ procedural and conceptual knowledge more than worked examples alone or problem solving.

**What does this look like in practice?**

Below is an example of four problems related to training thresholds. As you will see, the worked examples are faded to support learners in the first instance, but encourage greater independence as they develop their understanding.

Depending on the quality and understanding of learners’ responses on problem 2, they may be encouraged to move directly to problem 4, or continue with problem 3. This (sort of differentiation) is a demonstration of adaptive fading. Problem 4 requires more of a conceptual understanding as, unlike the other problems, the learners need to make their own decision about the threshold based on their understanding of the goal.

This approach can be used across a range of subjects and levels. We know that worked examples support novices, but if there are ‘relative’ experts in the classroom, be mindful of the expertise reversal effect and therefore, consider adaptive fading with the worked examples.

Once again, if I have misunderstood anything, feel free to let me know. If you have any examples you’d like to share, please leave them in the comments below.