> For the complete documentation index, see [llms.txt](https://docs.neurolabai.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.neurolabai.io/examples-and-use-cases/investigate-uncertainty.md). # Investigate Uncertainty Modafinil is the v0.1 uncertainty example because one part of its frozen report is visibly unstable while other parts are consistent. That makes a crucial point: reliability belongs to a route, head, field, and support context—not to the report as a single percentage. ## Read agreement per head | Head | Top item | Seed agreement | Interpretation | | -------------- | --------------------------------------- | -------------: | ---------------------------------------------------------- | | MoA | dopamine reuptake inhibitor (1.060) | 3/3, high | Stable among the three correlated members, not proven true | | Cognition | impulsivity (0.686) | 1/3, low | Demote top-1 emphasis; retain a shortlist | | Reported usage | ADHD label rank | 2/3, medium | Non-authoritative for CNS use under this release | | Side effect | Hypertension | 3/3, high | A sparse shortlist, not a safety assessment | | ATC | N06BA Centrally acting sympathomimetics | 3/3, high | Class hint; verify finer subcodes | The top cognition scores are nearly tied: `impulsivity` 0.686, `vigilance` 0.685, and `executive function` 0.680. A responsible UI or summary should present them as an unstable top-K set, not promote the first label as a confident winner. ![Modafinil Raw Payload workspace used to inspect exact per-head reliability fields](https://3908477117-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fgze1K0yofnhFNmM9nEyt%2Fuploads%2Fgit-blob-0e41b65f679bc79c4a68411a3df238746256016f%2F03-raw-payload%20\(1\).jpg?alt=media) ## Do not let a stable head erase another head's warning MoA agreement is high, but MoA graph coverage is low: only 3 binding/MoA-relevant edges against the threshold of 21. The ranking and the coverage warning answer different questions, so both remain visible. The Chat Report summary is useful for orientation, but it is not the end of the review. ![Modafinil Chat Report showing the top mechanism summary](https://3908477117-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fgze1K0yofnhFNmM9nEyt%2Fuploads%2Fgit-blob-ab83bbffb9825645771cedb41e5ddc2902a017c5%2F04-chat-report%20\(1\).jpg?alt=media) ## Audit source-derived text The FDA-derived usage array includes plausible sleep-related phrases such as `narcolepsy` and `obstructive sleep apnea`, but also generic extraction fragments such as `choice`, `continuous`, and `period`. This is an observed retrieval-quality limitation in the frozen response. The correct response is not to treat the entire array as validated because its records link to SPL documents. Verify the source section, keep provenance attached, and report extraction noise. Source identity and parsing quality are separate dimensions. ## Interpret the graph as a support artifact The Modafinil report-local graph contains 77 nodes, 85 edges, 85 paths, and 14 sources. It can help find the SLC6A3 binding context and related report entities, but it is not a literal account of how the GNN computed the top ranking. ![Modafinil report-local graph focused on the compound and accessible entity list](https://3908477117-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fgze1K0yofnhFNmM9nEyt%2Fuploads%2Fgit-blob-a7dad318e140c3cc206c1d40c8b88af5023ff8e2%2F07-knowledge-graph-detail%20\(1\).jpg?alt=media) ## Interpret the brain channel as diagnostic The default capture is again `Legacy specific map (diagnostic)` and carries the unvalidated legacy label. Relative Frontoparietal and other displayed values are not measurements and are not comparable across compounds as effect sizes. ![Modafinil Brain Explorer showing the rendered brain plus the legacy diagnostic limitation label](https://3908477117-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2Fgze1K0yofnhFNmM9nEyt%2Fuploads%2Fgit-blob-c057c5ec2dfa0f65b9ee8fb3418120487e9f159c%2F08-brain-localization-detail%20\(1\).jpg?alt=media) ## A defensible summary > In this pinned release, Modafinil's MoA ensemble was consistent across its members but had low KG coverage, while the cognition ensemble disagreed and produced a near-tied top three. The cognition result should therefore be presented as a shortlist, label-derived phrases require source-section review, and the brain display remains diagnostic rather than measured. Uncertainty has changed the prose, not merely added a footnote. That is the behavior Neurolab's reliability metadata is designed to support. *** **Fixture:** sanitized SHA-256 `beef1b5d0279e43dd350cde9db283bc6c852b74f32b2a7af4d060a237da8694f`. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.neurolabai.io/examples-and-use-cases/investigate-uncertainty.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.