Exemplar Based Feedback¶

Strategy · All strategies
Evidence · 2 claims (2 for) · 6 studies (3 quant-synthesis, 2 causal, 1 review), q2–q4 · 3 of 6 report an effect size

Description¶

Exemplar based feedback replaces (or supplements) verbal and rubric-only feedback with concrete samples of work — strong exemplars, weak exemplars, or a graded range — that learners compare against their own drafts. The learner's task shifts from decoding abstract criteria to judging the distance between their work and a visible target, which is how assessment standards are actually communicated in practice communities.

Design Implications¶

Exemplars work because they externalize quality standards that criteria lists leave tacit; learners consistently misinterpret rubric language until they see what "coherent argument" or "adequate evidence" looks like in a real artifact Formative feedback improves learning when learners understand quality standards. [+S]. The mechanism is comparative judgment: placing one's own work beside a model makes gaps salient in a way that generic comments do not. Comparing multiple exemplars — especially contrasting a strong and weak one — outperforms studying a single exemplar, because variation reveals which features carry the quality rather than the exemplar's surface content Comparing contrasting cases improves learning. [+S].

Context¶

Requirements¶

• Authentic exemplars at clearly differentiated quality levels, ideally annotated to explain why each is strong or weak
• An activity structure that forces comparison — learners must evaluate exemplars or self-assess against them, not just read them (Practice)
• A follow-up revision cycle in which learners apply the calibration to their own work (Assessment)

Constraints¶

• Copying: learners may imitate surface features (topic, structure, phrasing) rather than the underlying quality criteria, especially when only one exemplar is given [~M]
• Exemplars of exceptional quality can demotivate learners who see the gap as unbridgeable rather than as a target [~W]
• If exemplars are not annotated, learners often attend to irrelevant features; unguided exposure to models produces weak calibration [-W]
• Less effective for highly novel or creative tasks where no prior exemplar legitimately exists

Implementation Variability¶

• Graded range (ladder): three to five samples spanning quality levels; learners order and justify the ordering before seeing marks
• Contrasting pair: one strong and one weak exemplar on the same task, maximizing feature salience
• Exemplar + self-assessment: learners grade their own draft against exemplars before submitting, then receive confirmation or correction
• Progressive disclosure: full exemplar first, then partial exemplar with sections removed for learners to complete (Fading)

Target Learners¶

• Novices who lack the tacit standards of a discipline and cannot infer quality from criteria language alone [+S]
• Returning or first-generation students unfamiliar with academic work conventions
• Less beneficial for advanced learners who have already internalized quality standards; for them exemplars add little beyond the criteria themselves [~W]

Target Learning Goals¶

• Calibration of self-assessment: accurate judgment of one's own work quality
• Understanding of criteria and standards in complex, holistic tasks (essays, designs, code reviews)
• Transfer of quality concepts across tasks, via comparison of multiple cases

Instructions¶

1. Select 2–4 authentic exemplars on the target task, spanning a deliberate quality range; anonymize and, if needed, adapt to the current topic.
2. Annotate each exemplar to make the quality-relevant features explicit — or have learners do the annotating (Annotating).
3. Ask learners to evaluate the exemplars against the rubric before receiving marks, then discuss disagreements (Class Discussion).
4. Have learners compare their own draft to the exemplars and identify the largest gap (Self-Assessment).
5. Return to the exemplars after revision to check that the gap closed; repeat with harder exemplars as competence grows.

Related Strategies¶

• Use Worked Examples — the problem-solving analogue: exemplars of process rather than product
• Think-Aloud Modeling — makes the production of exemplar-quality work visible, complementing the product-level view
• Action-Oriented Feedback — combines well: exemplar shows the target, action comments specify the next step toward it

Examples¶

• Comparative judgement platforms such as No More Marking present learners and judges with pairs of authentic student work; the pairwise comparisons themselves constitute exemplar-based calibration at scale.
• Writing across the curriculum programs commonly distribute annotated A/B/C-range sample essays at the start of a term; students rank them and justify rankings before their first submission.
• Code review practice in software engineering courses uses exemplar pull requests — strong and flawed — that students critique before reviewing peers' code.

Key Sources¶

• Sadler, D. R. (1989). Formative assessment and the design of instructional systems. Instructional Science, 18(2), 119–144. doi:10.1007/BF00117714
• Black, P., & Wiliam, D. (1998). Assessment and classroom learning. Assessment in Education: Principles, Policy & Practice, 5(1), 7–74. doi:10.1080/0969595980050102
• Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. doi:10.3102/003465430298487
• van Gog, T., & Rummel, N. (2010). Example-based learning: Integrating cognitive and social-cognitive research perspectives. Educational Psychology Review, 22(2), 155–174. doi:10.1007/s10648-010-9134-7