# AI addiction policy review

I reviewed the first-pass HTML and its 12-person dataset on November 12, 2026. I assessed the policy framing and coding rules without independently validating every linked source.

1. **Readers need dates on the member cards before they infer current positions.** The cards summarize a 2013 veto, a 2020 interview, and recent campaign statements without displaying those dates together. I would show the year beside each summary and describe historical actions in the past tense, especially LePage’s veto and Barr’s recovery housing endorsement.

2. **Readers need to distinguish addiction records from related mental health records.** Hamawy has one mental health record and no addiction-specific record, yet his card reports one evidence record in the same format as addiction profiles. I would label the two counts separately and state his addiction review status directly on the card.

3. **Researchers need finer policy tags before they expand the dataset.** The broad overdose category combines naloxone access, fentanyl test strips, and supervised consumption proposals, while treatment access does not distinguish medications, workforce capacity, or treatment settings. I would add instrument and substance fields, including explicit categories for alcohol and stimulant policy, while leaving unresearched categories empty rather than assigning inferred positions.

4. **Researchers need to separate the documented action from the source describing it.** Talarico’s record reports coauthorship but uses the evidence type “Reported statement,” while Wiener’s sponsorship receives a different type despite also relying on an office account. I would store the action, source genre, and verification method separately so readers can compare sponsorship records without confusing primary and secondary documentation.

5. **Researchers need separate action dates and publication dates in their exports.** Talarico’s April 27 date identifies the news report, and Wiener’s June 30 date identifies an Assembly passage announcement rather than the original sponsorship event. I would preserve those dates as source dates and use a qualified event date or an explicit unknown when the source does not establish when the person acted.

6. **Readers need a plain-language explanation of the HALT Fentanyl Act beside the vote.** Pappas, Barr, and Letlow have recorded votes, but the page currently gives readers a bill title without describing the decision Congress considered. I would add a short, primary-source-grounded description of the bill’s scheduling and research provisions, then retain the warning that a whole-bill vote does not establish support for every provision.

7. **Researchers need a reproducible review plan before they infer party differences.** The pilot appropriately discloses selective inclusion, but available sources still determine which person appears with a historical objection, a supportive statement, or no record. I would log the same issue searches and review window for every member, add a member-by-issue search-status table, and postpone party-level policy summaries until reviewers complete that process.

I would preserve the separate election sources, the explicit distinction between signatures and endorsements, and the absence of policy scores. I would also retain the repeated warnings that missing evidence does not establish opposition.