PromptHelper: A Prompt Recommender System for Encouraging Creativity in AI Chatbot Interactions

TL;DR

PromptHelper introduces prompt recommender systems (PRS) to address users' difficulty in exploring prompt space and articulating intent while interacting with generative AI. The authors implement PromptHelper as an in-guide, context-aware prompt recommender embedded in a writing chatbot and evaluate it via a 2x2 within-subjects study across creative and academic writing tasks. Results show increased perceived exploration and, for academic writing, expressiveness without added cognitive load, suggesting PRS can expand creative exploration while preserving user agency. The work provides open-source resources and discusses broader implications for exploratory AI interfaces and AI literacy.

Abstract

Prompting is central to interaction with AI systems, yet many users struggle to explore alternative directions, articulate creative intent, or understand how variations in prompts shape model outputs. We introduce prompt recommender systems (PRS) as an interaction approach that supports exploration, suggesting contextually relevant follow-up prompts. We present PromptHelper, a PRS prototype integrated into an AI chatbot that surfaces semantically diverse prompt suggestions while users work on real writing tasks. We evaluate PromptHelper in a 2x2 fully within-subjects study (N=32) across creative and academic writing tasks. Results show that PromptHelper significantly increases users' perceived exploration and expressiveness without increasing cognitive workload. Qualitative findings illustrate how prompt recommendations help users branch into new directions, overcome uncertainty about what to ask next, and better articulate their intent. We discuss implications for designing AI interfaces that scaffold exploratory interaction while preserving user agency, and release open-source resources to support research on prompt recommendation.

Figures (4)

• Figure 1: We introduce prompt recommender systems (Figure \ref{['fig:teaser']}), a distinct class of prompt interaction methods. Existing approaches surface recommendations either as cold-start prompts or static categories at conversation entry (a–b), or as auto-complete (c) or opportunistic follow-ups appended to model responses (d). These designs primarily support prompt initiation or local refinement, but offer limited support for user-controlled exploration, branching, or sustained navigation of prompt space. PRS instead support exploration as an ongoing interaction, presenting prompts as persistent, selectable alternatives that users can compare and iteratively refine over time (e), treating prompts as recommender-eligible items.
• Figure 2: Across tasks, PromptHelper improved users’ perceived ability to explore and express ideas, while leaving workload and usability unchanged, suggesting that prompt recommender systems can scaffold exploratory interaction without imposing additional cognitive burden. Orange indicates ratings for the baseline chatbot, and blue indicates ratings for the baseline chatbot with PromptHelper enabled. We report the mean and standard error of all Likert scales using a 95% confidence interval.
• Figure 3: Example prompt item provided to the LLM, illustrating the structured fields (task, category, context, title, and prompt) used to define a recommendation.
• Figure 4: Privacy Policy Provided to Human Participants Recruited via the Prolific Platform.