AI & Technology|May 20, 2026|Francis John
When Should AI Interact First? The Accountability Loop for Proactive AI
When Should AI Interact First? The Accountability Loop for Proactive AI
Abstract
AI interaction is shifting from simple answer engines to proactive AI. But many proactive systems still follow an observer pattern: watch for a visible or programmatic signal, then react.
The Accountability Loop offers a different pattern. It starts with a shared goal. The agent utilizes the power of accountability to have the user commit to a plan, waits for time to pass, and follows up later. This turns proactivity from generic nudging into a commitment-based interaction pattern that better aligns with human psychology.
This pattern is important for AI systems that support long-horizon shared goals: improving health, managing finances, learning a skill, changing behavior, or coordinating care. At BodyBuddy, we have already been exploring this pattern in health accountability. With this grant, we will formalize the Accountability Loop as a public specification, reference implementation, and evaluation rubric for proactive AI agents.
Problem: The Trust Budget
Every time the AI speaks first, it spends some of the user’s attention. If the message feels irrelevant, too frequent, or poorly timed, the user learns to ignore it. There is a trust budget. Long-horizon goals make this harder because the agent may need to stay involved for weeks or months.
This is why it is critical to build proactive ai on a foundation where the trust budget is preserved.
Solution: The Accountability Loop
The Accountability Loop is an interaction pattern for proactive agents pursuing long-horizon shared goals across people, systems, and time.
The loop has seven parts:
- Establish the long-horizon goal.
- Infer what should happen next.
- Identify the accountable actor: the user, another person, an organization, or the agent itself.
- Create or confirm the next commitment.
- Wait for the relevant time or dependency to pass.
- Follow up with the right actor, take an allowed action, or escalate.
- Re-evaluate progress and repeat until the goal changes, completes, or no longer matters.
The Accountability Loop preserves the trust budget because the agent interacts first only when it is returning to a shared goal, prior commitment, or owned next step. Proactivity is no longer a random interruption; it is expected follow-through.
Research Agenda and Timeline
Over six months, I will build and publish a reference implementation of the Accountability Loop: a proactive agent pattern for long-horizon goals.
The work will use health accountability as the primary example, with additional lightweight examples in finance, learning, and care coordination to show that the pattern generalizes.
Timeline:
- Months 1-2: define the Accountability Loop and build the first working health accountability example.
- Months 3-4: generalize the implementation across additional long-horizon goal examples.
- Months 5-6: publish the reference implementation, evaluation rubric, and technical report.
Deliverables:
- Public Accountability Loop specification
- Reference implementation
- Example agents across several long-horizon goal domains
- Evaluation rubric
- Technical report: “When Should AI Interact First?”
Principal Investigator and Contributors
Principal Investigator: Francis John
Founder of BodyBuddy. Francis is leading the research, design, implementation, evaluation, and publication of the Accountability Loop work.
Expected contributors
No additional formal contributors are confirmed at this time. As part of the research process, I plan to conduct structured feedback conversations with professional accountability coaches, healthcare practitioners, and AI product builders who have experience supporting long-horizon behavior change. These conversations will be used to test whether the Accountability Loop reflects real accountability workflows and to identify failure modes around timing, trust, escalation, and follow-through.
CV attached: Francis John
Want daily accountability?
BodyBuddy texts you every day.
Build a healthier relationship with food and movement — one text at a time.
Join 500+ usersstaying healthy