Adding a second writing agent increased reported cognitive load, and writing confidence did not increase monotonically with support, with the Single condition reporting the lowest confidence¶

Claim · All claims
Evidence · 1 study · causal r? · q2
Currency · newest study 2026 · checked against its sources 2026-10-20 (automated, 1 of 1 entry) · not reviewed by a person

Subclaims¶

q2 i? Higher reports of feeling overwhelmed or confused during writing were most pronounced in the Duo condition, and the Single condition reported the lowest independent-writing confidence while Duo reported higher confidence than Single. → Harsh Kumar 2026

Evidence¶

Harsh Kumar 2026¶

Harsh Kumar, Zi Kang (Jace) Mu, Jonathan Vincentius, Ashton Anderson. (2026). Human Thinking under Plural LLM Assistance: Mathematical Problem Solving and Open-Ended Writing. Working Paper, July 2026, Toronto, Canada. https://arxiv.org/abs/2604.02677 doi:10.48550/arxiv.2604.02677

q2 · i? · causal · r?

Post-task self-report perceptions in Experiment-2 (confidence, helpfulness, idea stimulation, cognitive load) shown in Figure 8; findings are descriptive self-reports with no printed effect sizes, and participant quotes illustrate distraction from pop-up suggestions in the Duo condition.

"These perceived benefits were accompanied by higher reports of feeling overwhelmed or confused during writing, most pronounced in the Duo condition, suggesting that adding a second agent can increase cognitive load even when it provides additional inspiration."

Discussion¶

Learner Variables¶

• Working Memory — outcome: instruction changes it

Related Claims¶

• Mindful-condition students rated the pedagogical agent as more supportive and more concerned about their well-being on single items — related
• Single-LLM writing assistance increased idea-level homogenization across writers, while a role-specialized duo of architecturally distinct LLMs preserved baseline diversity — related
• Synchronous autonomous AI suggestions increased perceived writing efficiency but also caused distraction and mental overload during writing — related
• In math problem solving, a tutor-plus-peers LLM arrangement produced the strongest unassisted post-task test accuracy, with accuracy increasing monotonically from Control to Tutor + Peers — reports the opposite
• Training on AI interaction strategies slightly increases cognitive load during AI-assisted design — related
• Pedagogical agents as coaches and tutors are most useful with novice learners, while virtual role-players benefit learners across expertise levels — related
• An agentic cursor increased participants' understanding and predictability of the AI agent's actions during collaborative writing — related
• Synchronous AI suggestions shape subsequent writing, offering new ideas but risking influence over the writer's voice and argument — related
• In open-ended writing, both single-generalist and role-specialized duo LLM assistance improved essay quality over no-AI control, with no significant difference between them — related