Facilitating Asynchronous Idea Generation and Selection with Chatbots

TL;DR

The paper investigates how chatbots can enable asynchronous ideation and selection by translating human facilitation practices into automated agents. It designs two chatbots—a structured facilitator and an adaptive facilitator—supported by a semantic similarity classifier and a multi-armed bandit to guide ideas and adapt to user performance. Through two studies with 48 participants and an expert evaluation, both chatbots effectively fostered building on others’ ideas and converging toward promising concepts, though social interaction and accountability remained weak points compared with human facilitators. The findings advocate for a hybrid human–AI facilitation model to meet social and collaborative needs while providing continuous, on-demand guidance for asynchronous ideation. The work also contributes open resources, including code and prompts, to implement LLM-based chatbot facilitators in design contexts.

Abstract

People can generate high-quality ideas by building on each other's ideas. By enabling individuals to contribute their ideas at their own comfortable time and method (i.e., asynchronous ideation), they can deeply engage in ideation and improve idea quality. However, running asynchronous ideation faces a practical constraint. Whereas trained human facilitators are needed to guide effective idea exchange, they cannot be continuously available to engage with individuals joining at varying hours. In this paper, we ask how chatbots can be designed to facilitate asynchronous ideation. For this, we adopted the guidelines found in the literature about human facilitators and designed two chatbots: one provides a structured ideation process, and another adapts the ideation process to individuals' ideation performance. We invited 48 participants to generate and select ideas by interacting with one of our chatbots and invited an expert facilitator to review our chatbots. We found that both chatbots can guide users to build on each other's ideas and converge them into a few satisfying ideas. However, we also found the chatbots' limitations in social interaction with collaborators, which only human facilitators can provide. Accordingly, we conclude that chatbots can be promising facilitators of asynchronous ideation, but hybrid facilitation with human facilitators would be needed to address the social aspects of collaborative ideation.

Figures (12)

• Figure 1: A conceptual model of synchronous and asynchronous ideation. Unlike synchronous ideation where collaborators (C1 and C2) develop ideas by taking turns, we focus on asynchronous ideation where collaborators generate ideas individually, at their own pace and method (right). For example, while C1 diversifies ideas, C2 could focus on improving ideas. Here, facilitators' role will be presenting C1's idea as an inspiration to C2.
• Figure 2: We designed chatbots that interact with individual collaborators and facilitate their asynchronous idea generation (left) and selection (right). We adapted guidelines from the literature on human facilitators to design our chatbots' behaviors, presenting other collaborators' ideas as inspiration and suggesting ideation methods to guide individuals to build on each other's ideas.
• Figure 3: Collaborative idea generation (top) and selection (down) led by facilitators. By showing other group members' ideas and opinions, facilitators can guide individual collaborators to effectively generate more ideas and evaluate the ideas with more diverse viewpoints. We designed chatbot facilitators that could guide such effective ideation in asynchronous settings.
• Figure 4: During idea generation (left), the chatbots repeat the cycle of presenting other group members' ideas as inspiration (a), suggesting an ideation method (b), and requesting ratings on the users' own ideas (c). Then, the chatbots continue showing other types of inspiration and ideation methods (d). During idea selection (right), the chatbots show a collected idea (e), request users' initial opinion (f), show other group members' opinions (g), suggest re-evaluating the idea (h), and request ratings on the idea.
• Figure 5: The system structure of the adaptive facilitator. During idea generation (top), the MAB system selects inspirations and ideation methods based on the individuals' rating of their own ideas. During idea selection (bottom), the MAB system presents ideas based on the users' collective rating of each idea (i.e., prioritizing ideas with uncertain group opinions).