ROSE – A Framework for Enhancing Human–AI Communication through Structured Prompting
Introduction – why structure still matters
Generative AI is already reshaping higher education (Lee et al., 2024), software engineering (Coutinho et al., 2024) and customer-service operations (McKinsey & Company, 2023a). Yet every practitioner quickly discovers the garbage-in–garbage-out principle: vague prompts yield vague answers (Homolak, 2023). This has created a lively new discipline—prompt engineering—now recognised by consultancies such as McKinsey as a strategic workforce skill (McKinsey & Company, 2023b) and even advertised as a six-figure career path (Kelly, 2024).
Against a backdrop of increasingly elaborate frameworks (e.g., Chain-of-Thought, ReAct), the ROSE model—developed by Ian Farmer—offers a deliberately simple mnemonic focused on four essentials:
| R | O | S | E |
|---|---|---|---|
| Role | Objective | Style | Exemplar |
Its purpose is to give everyday users a checklist they can recall in seconds yet still embody best practice drawn from the literature on persona setting (Wei et al., 2022), goal specification (Yao et al., 2022), output formatting (Zhang et al., 2024) and few-shot learning (Brown et al., 2020).
Deconstructing ROSE
| Component | Purpose | Typical prompt language | Why it works |
|---|---|---|---|
| Role | Assigns a persona/expertise lens | “Act as a financial-times columnist…” | Primes the model’s tone and knowledge domain (Forward Future, 2025) Forward Future |
| Objective | Defines the exact task or deliverable | “…draft a 250-word op-ed on UK carbon markets.” | Clear goals reduce irrelevant digressions (Yao et al., 2022) |
| Style | Specifies tone, format, length | “…use British English, bullet points, APA citation.” | Matches audience expectations; aids downstream integration (Zhang et al., 2024) |
| Exemplar | Supplies positive/negative examples | “Here is a model paragraph; avoid clichés like…” | Leverages few-shot pattern learning (Brown et al., 2020) |
These four elements interact: changing Role (e.g., “friendly tour-guide”) inevitably alters suitable Style, and an Exemplar often clarifies both content depth (Objective) and layout (Style).
Why ROSE improves results
Systematic thinking – ROSE provides a cognitive scaffold that forces users to plan before they prompt, a technique shown to improve student prompt-writing efficacy (Nguyen et al., 2024).
Greater alignment – Explicit role + objective + style constraints cut hallucination rates and raise factual accuracy (Zhang et al., 2024).
Efficiency – Better first-attempt prompts reduce iterative re-prompting, a key pain-point reported by novice users (Impromptu Framework authors, 2025).
AI-literacy builder – ROSE’s simplicity makes it a useful entry point for classroom AI-literacy programmes (Digital Promise, 2024).
ROSE in action
Scenario 1 – Blog-post outline
Basic prompt
“Write a blog-post outline on remote work and well-being.”
ROSE prompt
Role: HR consultant specialising in employee wellness.
Objective: Draft a 5–7 section outline for HR managers showing mental and physical health benefits of remote work, with data examples.
Style: Professional yet engaging; nested bullet list.
Exemplar: Positive – “Section: Autonomy → cite study linking autonomy & job satisfaction.”
Negative – “Avoid generic ‘work-life balance’ claims without evidence.”
Expected difference – The ROSE version delivers targeted, evidence-rich headings formatted for immediate publication, whereas the basic prompt tends to produce generic platitudes.
Scenario 2 – Research briefing
Basic prompt
“Summarise AI in education research.”
ROSE prompt
Role: Research assistant preparing a 500-word briefing for university staff.
Objective: Synthesis of peer-reviewed studies (2022-2024) on generative AI in HE assessment; include actionable insights.
Style: Neutral academic tone; thematic bullet headings.
Exemplar: Positive – cite finding on redesigning assessments for higher-order skills;
Negative – exclude vendor marketing.
Dialogical
It’s also important for students to not presume that AI prompting dialogues are the same as search. Searching is a hunt for a fixed answer. It’s transactional, like casting a fishing line and hoping for a bite. You type a query, you get a link.
Seeking, by contrast, is an open-ended, iterative dialogue. It involves exploring, re-framing questions, and following unexpected threads of meaning. In an AI context, seeking invites discovery, not just the retrieval of a pre-existing fact. It’s a dialogical conversation.
This “seeking” mindset is precisely what we want to nurture in students, especially when they face ambiguity and complexity (Carless & Boud, 2018; Xu, 2025). Helping students navigate ambiguity is a crucial life skill, particularly in fields like sustainability and ethical decision-making, where clear-cut answers are rare.
Conclusion
ROSE’s chief virtue is memorability. With these four words—Role, Objective, Style, Exemplar—users gain a repeatable template that captures the core insights of much denser prompt-engineering research. In practice it delivers clearer intent, sharper outputs and faster workflows while doubling as an accessible learning aid for the wider workforce. As LLMs become ever more capable, the ability to articulate our needs with equal clarity will only grow in importance; ROSE is a simple, robust step in that direction.
References
Brown, T. B., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., … (2020). Language models are few-shot learners [arXiv:2005.14165].
Coutinho, M., Marques, L., Santos, A., Dahia, M., França, C., & Santos, R. (2024). The role of generative AI in software development productivity: A pilot case study [arXiv:2406.00560].
Digital Promise. (2024). AI literacy: A framework to understand, evaluate and use emerging technology. digitalpromise.dspacedirect.org
Forward Future. (2025, April). The “Act As” role-based prompting. Forward Future
Homolak, D. (2023). “Garbage in – garbage out!” Still applies with Gen AI. Integreon Insights. Integreon –
Kelly, J. (2024, March 6). The hot, new high-paying career is an AI prompt engineer. Forbes. Forbes
Lee, D., Arnold, M., & Srivastava, A. (2024). The impact of generative AI on higher education learning and teaching: A study of educators’ perspectives. Computers & Education: X, 4, 100115. ScienceDirect
McKinsey & Company. (2023a). The state of AI in 2023: Generative AI’s breakout year. McKinsey & Company
McKinsey & Company. (2023b). What is prompt engineering? McKinsey Explainers. McKinsey & Company
Nguyen, T., Chen, L., & Patel, S. (2024). Effects of a prompt-engineering intervention on undergraduate AI self-efficacy. [arXiv:2408.07302].
Wei, J., Wang, X., Schuurmans, D., Bosma, M., Chiang, C. A., … (2022). Chain-of-thought prompting elicits reasoning in large language models [arXiv:2201.11903].
Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., … (2022). ReAct: Synergising reasoning and acting in language models [arXiv:2210.03629].
Zhang, H., Li, P., & Sun, X. (2024). Does prompt formatting have any impact on LLM performance? [arXiv:2411.10541].