CowPilot: A Framework for Autonomous and Human-Agent Collaborative Web Navigation

TL;DR

CowPilot is a framework that enables effective human-agent collaboration in web navigation, significantly reducing human effort while maintaining high task success rates through flexible interaction modes.

Contribution

It introduces a novel collaborative web navigation framework allowing seamless human-agent interaction and evaluates its effectiveness across multiple websites.

Findings

Achieves 95% task success rate with minimal human input

Humans perform only 15.2% of total steps on average

Agent can complete up to half of tasks independently

Abstract

While much work on web agents emphasizes the promise of autonomously performing tasks on behalf of users, in reality, agents often fall short on complex tasks in real-world contexts and modeling user preference. This presents an opportunity for humans to collaborate with the agent and leverage the agent's capabilities effectively. We propose CowPilot, a framework supporting autonomous as well as human-agent collaborative web navigation, and evaluation across task success and task efficiency. CowPilot reduces the number of steps humans need to perform by allowing agents to propose next steps, while users are able to pause, reject, or take alternative actions. During execution, users can interleave their actions with the agent by overriding suggestions or resuming agent control when needed. We conducted case studies on five common websites and found that the human-agent collaborative mode achieves the highest success rate of 95% while requiring humans to perform only 15.2% of the total steps. Even with human interventions during task execution, the agent successfully drives up to half of task success on its own. CowPilot can serve as a useful tool for data collection and agent evaluation across websites, which we believe will enable research in how users and agents can work together. Video demonstrations are available at https://oaishi.github.io/cowpilot.html