Skip to content

Study 001: MCP clinic interface case study

Exploratory pilot concluded · updated 15 August 2026

Study 001 · Coordinator-reviewed case study

We tested Human Standards guidance on nine AI-built clinic apps.

We generated nine clinic interfaces with gpt-5.6-sol under three guidance conditions. Directed MCP access produced the strongest automated medians; the coordinator still found obvious keyboard, focus, validation and state-transition problems in every condition.

9generated interfaces
3guidance conditions
8critical automated scenarios
0independent reviews completed

Method

Three treatments, three recorded runs each

All nine interfaces were generated with Codex CLI 0.147.0 using gpt-5.6-sol at medium reasoning effort and the default service tier. The CLI was authenticated through a ChatGPT account. The model, client, system instructions and ordinary tools were fixed; only the guidance condition changed.

Runs were isolated, internet access was disabled, the mock API was deterministic, and the browser clock and timezone were fixed. Each interface had to book, validate, recover, reschedule and cancel; work by keyboard; and remain usable at mobile width. We did not repair or selectively rerun model-produced defects.

A

Baseline

Specification only

B

Directed MCP

Prompted to retrieve and apply Human Standards through MCP

C

Frozen context

The same project received a static Human Standards reference packet

All three MCP runs actually used the server. They made 7, 5 and 11 completed retrieval calls respectively across search, standards and heuristics tools.

Automated evidence

The MCP condition led on both automated measures

Higher critical completion and lower serious accessibility counts are better. With three runs per condition, these medians describe this sample and do not estimate a general effect.

A Baseline
Median critical completion
50%
range 25–87.5%
Median serious rule–state pairs
8
range 7–11
B Directed MCP
Median critical completion
87.5%
range 50–100%
Median serious rule–state pairs
4
range 3–4
C Frozen context
Median critical completion
87.5%
range 87.5–87.5%
Median serious rule–state pairs
7
range 6–11
Every recorded run; no outputs removed
Condition Artifact Run Critical scenarios Serious rule–state pairs Affected nodes Unreachable axe states
A V006 R002 4/8 7 16 3 states
A V009 R006 7/8 11 202 None
A V007 R008 2/8 8 16 3 states
B V004 R001 8/8 4 15 None
B V003 R004 7/8 3 13 None
B V002 R009 4/8 4 8 3 states
C V008 R003 7/8 6 19 None
C V001 R005 7/8 7 22 None
C V005 R007 7/8 11 63 None

An unreachable state was reported as missing, never converted into a zero-violation accessibility pass. “Affected nodes” can be concentrated repetitions of the same rule; for example, V009’s 202 nodes should not be read as 202 distinct accessibility rules.

Coordinator walkthrough

Interaction problems appeared in all three conditions

The project coordinator viewed all nine artifacts while their condition labels were withheld and recorded qualitative observations. This was not the planned independent review and no rubric score was calculated.

01

Keyboard focus failed across conditions

Appointment and time cards repeatedly lacked usable keyboard focus. Some buttons and cards also lacked visible focus or hover states.

02

Visual polish did not predict task completion

V004 looked polished and passed all eight critical scenarios, but its radio cards were still difficult to use by keyboard. V007 looked coherent but completed only two critical scenarios.

03

Validation and transitions were often unclear

One baseline interface placed a missing-name error at the bottom of the form rather than beside the field. Another output failed to return the reader to the top when moving to review. One rescheduling path lacked a clear success state.

04

Several runs handled structure and state well

V003 retained entered data across backwards navigation and organised its confirmation well. V005’s home page was notably simple. V008 produced a streamlined confirmation. V009’s landing page and progress structure made the flow easy to understand.

Clinic details form showing an Enter your full name error below the entire form while the empty full-name field remains unmarked near the top
V006 · Baseline: the validation message is visible, but separated from the field that needs attention. The same frozen artifact was replayed after the walkthrough to capture the observed layout.

Across all nine

Every run failed refresh safety and triggered serious accessibility findings

Every run built and typechecked, and the browser recorded no external requests. But all nine failed the non-critical refresh-safety scenario: a confirmed booking was lost on refresh.

Every run also produced at least three serious accessibility rule–state pairs. Colour contrast was the recurring automated rule; isolated outputs also produced a definition list or unnamed-dialog problem.

Protocol change

Why we stopped before independent review

The preregistered plan required two independent reviewers to score all nine opaque artifacts before the condition mapping was revealed. We stopped after the coordinator walkthrough, revealed the mapping and closed recruitment because the visible cross-condition problems made this pilot more useful as a development set than as a first human-rated benchmark.

What this means: the preregistered P3 human-centred quality outcome is missing. The coordinator observations are not independent scores, so this case study cannot support a confirmatory claim about MCP impact.

What happens next

What we will change in Human Standards MCP

1

Make interaction guidance testable

Pair keyboard, focus and selection guidance with concrete implementation contracts and checks.

2

Retrieve by product task

Give agents a focused recipe for forms, commitment, failure recovery and responsive state transitions.

3

Require a self-check

Ask the agent to exercise keyboard paths, validation, retry states and truthful confirmation before finishing.

4

Validate on unseen work

Use these nine outputs as a development set, then test a frozen MCP version on a new task with blinded independent reviewers.

Inspect the evidence

Download the case data and frozen evidence

Each run identifier in the results table links to its frozen automated summary. Detailed functional, axe, focus and guardrail JSON files use the same directory and filename pattern. Screenshots show fictional data only.