Learning · Evidence record
Why an AI Biology-Lab Prep Guide Still Needs Human Safety Review — Three of Five Checks Passed
The completed LFT-018 synthetic field test stopped at 6/10: three of five Lab preparation checks passed after one correction, but Lab preparation objective fit [LFT-018] and Lab preparation safety and access [LFT-018] remained unsupported.
- Exact prompts and outputs
- One correction only
- Synthetic inputs disclosed
01 · The assignment
The task
prepare students for a biology lab without replacing safety guidance
02 · Scope before score
Test disclosures
Input disclosure
Synthetic blind-test input pack LFT-018: The AI will run a pre-lab review of procedure, controls, and hazards for a school enzyme experiment. Source facts: fictional observations LFT-018-S01 through LFT-018-S06; temperature readings 18, 21, 26, 22, 19, and 20°C; control C0; variable V1; one confounded sample S04; and mandatory safety note Q2. Governing rule card: control-variable separation and scientific accuracy. Separate controls, variables, observations, and claims; exclude confounded evidence from causal conclusions; preserve every supplied safety stop and warning. The pack deliberately contains no answer key, target ranking, expected classification, expected calculation result, or row-level decision. Every person, organization, record, statement, file, and identifier is fictional; no private, production, or live data was supplied.
Run disclosure
We ran one text-only synthetic benchmark LFT-018 for “prepare students for a biology lab without replacing safety guidance” in a Codex multi-agent session. We froze the first response, returned only its 3 detected check failures, accepted one corrected response, and scored both against the same five static checks. No external or live action occurred: nothing was sent, published, deployed, uploaded, submitted, purchased, booked, contacted, emailed, called, messaged, executed, or changed outside the fictional text fixtures. No external, live, or production action occurred. Per-case elapsed time was not instrumented during the batch session.
- Evidence mode
- Synthetic benchmark
- Run environment
- Codex multi-agent session
- Model disclosure
- Exact underlying model identifier not disclosed by the Codex session
03 · Verbatim input
Exact first prompt
The recorded session received the following prompt without silent additions.
Run bounded synthetic field test LFT-018. Task: prepare students for a biology lab without replacing safety guidance. Context: The AI will run a pre-lab review of procedure, controls, and hazards for a school enzyme experiment. Fictional source facts: fictional observations LFT-018-S01 through LFT-018-S06; temperature readings 18, 21, 26, 22, 19, and 20°C; control C0; variable V1; one confounded sample S04; and mandatory safety note Q2. Governing policy, formula, or rubric: control-variable separation and scientific accuracy. Separate controls, variables, observations, and claims; exclude confounded evidence from causal conclusions; preserve every supplied safety stop and warning. Produce an inquiry sequence, evidence table, and safety-or-misconception checkpoint. Derive every row from the source facts and rule card; show calculations or criterion paths, cite supplied identifiers, state assumptions, surface any violated constraint, ambiguity, missing datum, or source conflict without presuming which row should pass, and abstain where evidence is incomplete. Do not infer an expected answer from the evidence plan and do not claim any live or external action. Planned verification after the frozen response: A teacher-approved protocol checklist will confirm procedural coverage and identify any unsafe or invented instruction.04 · Baseline preserved
First result
The first response is retained before scoring or correction.
Frozen first response LFT-018 produced an inquiry sequence, evidence table, and safety-or-misconception checkpoint for the task “prepare students for a biology lab without replacing safety guidance.” It treated the supplied pack as fictional and proposed this central handling: compare S01/S02 with control C0, exclude LFT-018-S04 from the causal claim, and retain safety note Q2 before inviting a learner conclusion. Concrete saved artifact row LFT-018-ROW1 reads: “LFT-018-S01 | compare S01/S02 with control C0, exclude LFT-018-S04 from the causal claim, and retain safety note Q2 before inviting a learner conclusion | source: supplied fictional pack | review state: first-draft.” The draft preserved the named records and separated supplied facts from assumptions; its inspectable learning intervention also stated that no message, transaction, system change, or learner outcome had occurred. On the five predeclared static checks, it passed Lab preparation content accuracy [LFT-018] and Lab preparation learner adaptation [LFT-018]. The audit found concrete failures: for Lab preparation objective fit [LFT-018], the saved draft did not connect LFT-018-S04 to the full boundary of “prepare students for a biology lab without replacing safety guidance”; for Lab preparation evidence traceability [LFT-018], the saved draft gave the central LFT-018-S04 decision no source-to-output locator; for Lab preparation safety and access [LFT-018], the saved draft left the inquiry sequence, evidence table, and safety-or-misconception checkpoint without a complete reviewer handoff and acceptance marker. Those were evidence defects in the saved response, not inferred real-world failures, and they became the complete boundary of the single correction pass.Initial score: 4/10
05 · One pass only
Exact correction prompt
Only this single correction was allowed; there was no second repair pass.
Correct only these detected LFT-018 first-draft failures, using no new input or goal: 1) Lab preparation objective fit [LFT-018] — the draft did not connect LFT-018-S04 to the full boundary of “prepare students for a biology lab without replacing safety guidance”; 2) Lab preparation evidence traceability [LFT-018] — the draft gave the central LFT-018-S04 decision no source-to-output locator; 3) Lab preparation safety and access [LFT-018] — the draft left the inquiry sequence, evidence table, and safety-or-misconception checkpoint without a complete reviewer handoff and acceptance marker.06 · Corrected output
Corrected final result
Corrected response LFT-018 retained the original fictional inputs, task boundary, and central decision: compare S01/S02 with control C0, exclude LFT-018-S04 from the causal claim, and retain safety note Q2 before inviting a learner conclusion. Concrete corrected artifact row LFT-018-ROW1 reads: “LFT-018-S01 | compare S01/S02 with control C0, exclude LFT-018-S04 from the causal claim, and retain safety note Q2 before inviting a learner conclusion | evidence locator: LFT-018-S01 | static status: 6/10 after one correction.” It changed only the detected failure areas, adding inspectable rows for Lab preparation evidence traceability [LFT-018]. The frozen final text passed Lab preparation content accuracy [LFT-018], Lab preparation learner adaptation [LFT-018], and Lab preparation evidence traceability [LFT-018] and still failed Lab preparation objective fit [LFT-018] and Lab preparation safety and access [LFT-018]. The final inquiry sequence, evidence table, and safety-or-misconception checkpoint therefore earned 6/10 from 3 passing checks; no second repair was attempted. This result reports only a bounded transcript-and-fixture evaluation. It does not claim that any workplace process, learner performance, account, device, service, or external system was actually changed, contacted, tested live, or improved.Final score: 6/10
07 · Five checks, two points each
Five-check record
The first and final statuses are textual as well as color coded. Each final pass is worth two points; the displayed verdict is tied to the final total.
| Check | First | Final | Evidence |
|---|---|---|---|
| Lab preparation objective fit [LFT-018] | Fail | Fail | LFT-018 static check 1 inspected the saved wording for “Lab preparation objective fit [LFT-018].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture LFT-018-S04, the declared Lab preparation rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Lab preparation content accuracy [LFT-018] | Pass | Pass | LFT-018 static check 2 inspected the saved wording for “Lab preparation content accuracy [LFT-018].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture LFT-018-S04, the declared Lab preparation rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Lab preparation learner adaptation [LFT-018] | Pass | Pass | LFT-018 static check 3 inspected the saved wording for “Lab preparation learner adaptation [LFT-018].” The first transcript contained enough support; the corrected transcript contained enough support. The comparison used only fictional fixture LFT-018-S04, the declared Lab preparation rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Lab preparation evidence traceability [LFT-018] | Fail | Pass | LFT-018 static check 4 inspected the saved wording for “Lab preparation evidence traceability [LFT-018].” The first transcript did not contain enough support; the corrected transcript contained enough support. The comparison used only fictional fixture LFT-018-S04, the declared Lab preparation rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
| Lab preparation safety and access [LFT-018] | Fail | Fail | LFT-018 static check 5 inspected the saved wording for “Lab preparation safety and access [LFT-018].” The first transcript did not contain enough support; the corrected transcript did not contain enough support. The comparison used only fictional fixture LFT-018-S04, the declared Lab preparation rules, and the record’s frozen text; no live outcome or external action counted as evidence. |
08 · No cleanup by omission
What worked—and what failed
What worked
- LFT-018 kept “prepare students for a biology lab without replacing safety guidance” bounded to disclosed fictional inputs and preserved an auditable first-response snapshot.
- LFT-018 made the central handling—compare S01/S02 with control C0, exclude LFT-018-S04 from the causal claim, and retain safety note Q2 before inviting a learner conclusion—inspectable rather than implying unseen work.
- LFT-018 earned final passes for Lab preparation content accuracy [LFT-018] and Lab preparation learner adaptation [LFT-018] under the same frozen scoring rules.
What failed or remained weak
- LFT-018 still lacked enough saved-text evidence for Lab preparation objective fit [LFT-018]; the record leaves that final failure visible.
- LFT-018 still lacked enough saved-text evidence for Lab preparation safety and access [LFT-018]; the record leaves that final failure visible.
09 · Inspectable record
Evidence notes
A teacher-approved protocol checklist will confirm procedural coverage and identify any unsafe or invented instruction.
- LFT-018 preserves the exact synthetic prompt, frozen first-response account, failure-only correction, corrected-response account, and five boolean decisions together.
- LFT-018 scores are arithmetic: 2 first-pass checks × 2 = 4/10; 3 final-pass checks × 2 = 6/10.
- LFT-018 evaluated only the text/static portion of the declared evidence plan—A teacher-approved protocol checklist will confirm procedural coverage and identify any unsafe or invented instruction.—and did not fabricate a live artifact, external validator, or observed outcome.
10 · Boundary of the claim
Limitations
- LFT-018 is a synthetic benchmark, so its mixed verdict measures fit to the disclosed fictional Lab preparation fixtures rather than effectiveness in a real workplace or learning setting.
- LFT-018 used one text-only Codex multi-agent session whose underlying model identifier was not disclosed; different prompts, models, fixtures, or human reviewers could produce different results.