Reactive and proactive AI writing agents on a minimalist multimodal writing platform

Design · All designs
Evidence · 2 claims (2 for) · 1 study (1 causal), q2 · 0 of 1 report an effect size · 2 claims rest on one study

Description

The study developed two AI agents that manipulated conversational initiative while holding other features constant: both used the same interface, visual inputs, VizChat architecture, Retrieval-Augmented Generation pipeline, visualisation knowledge base, and multimodal GPT-4o model. The reactive agent generated only when the learner messaged; the proactive agent followed an expert-authored conversational state sequence, initiating orientation and ordered guiding questions per visual and giving confirmatory, clarifying, or corrective feedback grounded in visual evidence. The platform comprised a display area, a writing space, and a condition-specific chat or instructions component.

Design Implications

Context

Requirements

• Requires the same underlying model, retrieval pipeline, and task materials across conditions so the contrast isolates who initiates and sequences support

Constraints

• The proactive sequence's triggering events, question order, target knowledge, and completion rules were fixed across participants; the prompt prohibited asking about non-displayed visuals and repeating questions

Target Learners

• medical and nursing students interpreting healthcare simulation visualisations

Learning Goals

• multimodal analytical writing from data visualisations

Claims

• Proactive Agent Dialogues Integrated Epistemic Reasoning [+M]
• Proactive Engagement Constructive Adequate Reasoning Link [+M]

Related Designs

Examples

-

Key Sources

• Yueqiao Jin, Kaixun Yang, Roberto Martinez-Maldonado, Dragan Gašević and Lixiang Yan. (2025). The Agency Gap in AI-Supported Writing: How Reactive and Proactive Agent Designs Shape Multimodal Reasoning. https://arxiv.org/abs/2507.04398

Page metadata

Field	Value
type	design
id	reactive-proactive-ai-writing-agents-platform
title	Reactive and proactive AI writing agents on a minimalist multimodal writing platform
description	The study developed two AI agents that manipulated conversational initiative while holding other features constant: both used the same interface, visual inputs, VizChat architecture, Retrieval-Augmented Generation pip...
status	review
generated	by process:settle-candidates on 2026-10-12
trust	unverified — no one has confirmed this page's content
sources	1 — yueqiao-jin-2025

Frontmatter is the page's machine-readable record — see the Schema & Guide.