arXiv:2606.31980v1 Announce Type: new Abstract: Agents are increasingly capable of automating software tasks, but can they teach humans how to use software themselves? We introduce DigitalCoach, a multimodal dataset of 72 human expert-novice computer use coaching sessions consisting of 22,752 dialogue turns grounded in 28.1 hours of screen and input event recordings across five software applications. We use DigitalCoach to evaluate whether state-of-the-art models can teach humans how to use computers. Automated evaluation shows that models differ from humans in how they coach: models provide m
The proliferation of advanced AI models makes their application to complex human-computer interaction, such as coaching, a logical next step, with current research focusing on closing the gap between human and AI coaching effectiveness.
This research highlights a critical area where AI agents can extend their capabilities beyond automation to actively teach and upskill humans in software use, impacting future workforce training and individual productivity.
The understanding of AI's current limitations and potential in human-like coaching, specifically revealing where AI models diverge from human experts in educational strategies and grounding conversational context.
- · AI development platforms
- · Ed-tech companies
- · Software training providers
- · Productivity software developers
- · Traditional human-only software coaches
- · Companies with inefficient software onboarding
- · Basic tutorial providers
AI agents begin to effectively teach complex software functions to human users, improving digital literacy and efficiency.
The role of human instructors shifts towards higher-order problem-solving and curriculum development, while AI handles repetitive coaching tasks.
Personalized, on-demand AI coaches become ubiquitous, democratizing access to expert-level software proficiency and accelerating skill acquisition across diverse populations.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL