Towards Cognitive Process-Aware Proactive Writing Support
A research paper proposes a framework for proactive writing support based on Flower and Hayes' cognitive process theory of writing. The framework identifies 14 writing support types associated with six cognitive processes and characteristic interaction behaviors. The approach is instantiated in a system called AToM CoWriter. Two within-subjects studies with 21 participants provide initial evidence that the approach improves expressiveness and idea exploration, and that cognitive process inference increases engagement with proactive suggestions.
Large language models can support writing, but existing tools require users to explicitly articulate prompts, which is burdensome in creative writing where intentions are often ambiguous. Proactive support that infers users' needs from writing interactions could alleviate this burden. This work focuses on determining what support to provide by using Flower and Hayes' cognitive process theory of writing as an interpretable bridge between observable writing behavior and appropriate support types. Through a formative study and literature review, 14 writing support types associated with six cognitive processes are identified, along with characteristic interaction behaviors linked to each process. The framework is instantiated in AToM CoWriter, which infers support needs from writing interactions and document context. Two within-subjects studies (N = 21) provide initial evidence that this approach improves expressiveness and idea exploration, and that cognitive process inference increases engagement with proactive suggestions.
The system maps observable writing interactions to cognitive processes (e.g., planning, translating, reviewing) and then to specific support types. This requires modeling of interaction behaviors and document context to infer the user's current cognitive process. The approach leverages a theory-driven taxonomy rather than purely data-driven methods, potentially improving interpretability and alignment with user needs. The studies suggest that cognitive process inference can increase user engagement with proactive suggestions, indicating that timing and relevance of interventions are critical.
This research addresses a gap in AI writing tools, which are largely reactive and prompt-based. Proactive support could differentiate products in the competitive writing assistant market. The focus on creative writing suggests potential applications in content creation, marketing, and education. The finding that cognitive process inference increases engagement implies that users may be more receptive to suggestions that align with their current mental state, which could inform product design for user retention and satisfaction.
The framework could enable new features in AI writing assistants that proactively offer support, potentially increasing user engagement and productivity. For companies developing writing tools, this could lead to higher user retention and differentiation. The research may also inform the design of enterprise writing solutions where reducing cognitive load is valuable. However, the business value is currently at an early research stage, with no direct evidence of commercial adoption or revenue impact.
Next signals to watch include: (1) publication of the full paper with detailed methodology and results; (2) follow-up studies with larger and more diverse participant pools; (3) integration of the framework into commercial writing tools; (4) exploration of other cognitive theories or domains; (5) evaluation of long-term effects on writing quality and user trust. The approach may also be extended to other creative tasks such as design or coding.