Three-segment prompt design: Condition, Consequence, and Refinement¶

Strategy · All strategies
Evidence · 3 claims (2 for, 1 mixed) · 1 study (1 design), q2 · 0 of 1 report an effect size · 3 claims rest on one study

Description¶

The article describes a systematic way to write prompts for LLM-based feedback by splitting each prompt into three segments. The Condition specifies when support should be given (e.g., "If the students' answer is incomplete or incorrect"), the Consequence requests a specific output adaptable by the teacher and adaptive to the student's answer, and the Refinement includes a restriction such as "without giving the correct answer". This structure is intended to allow systematic manipulation and empirical investigation of adaptive and adaptable prompt segments.

Design Implications¶

Context¶

Requirements¶

• The teacher or researcher must specify the condition, the desired output type, and a refinement restriction for each prompt

Constraints¶

• The refinement restriction is included because prior research cited by the authors indicates that elaborated feedback combined with the correct answer has low efficiency and may lead to superficial learning

Target Learners¶

• teachers and researchers designing LLM-based feedback tasks

Target Learning Goals¶

• formative feedback design
• self-regulated learning

Claims¶

• Feedback quality in LLM-generated lessons improved with three or more segments compared to one- or two-segment prompting [+M]
• Task decomposition into three segments yields the highest-rated LLM-generated tutor training lessons, while single-step generation yields the lowest [+M]
• The highest decomposition level (five segments) performed slightly worse than the three-segment approach, suggesting non-monotonic benefits of decomposition [~M]

Related Strategies¶

• Leap Llm Formative Feedback Platform

Examples¶

-

Key Sources¶

• Steinert, S., Avila, K. E., Ruzika, S., Kuhn, J., & Küchemann, S. (2024). Harnessing large language models to develop research-based learning assistants for formative feedback. Smart Learning Environments. https://doi.org/10.1186/s40561-024-00354-1