Source-Backed Journal Reads: Build an Evidence Card

Put the existing evidence concepts to work

This worksheet puts the Pattern Receipt at /learn/glossary/pattern-receipt and Source Grounding at /learn/glossary/source-grounding into practice. The pattern detection guide at /learn/how-ai-journal-pattern-detection-works explains sources, thresholds, and uncertainty in more detail.

Daylogue's pattern feature reads mood, energy, stress, and sleep hours together across check-ins. The current pattern product page says those observations appear in plain language, but it does not promise a citation interface for every read. This card supplies a manual inspection method.

The anatomy of an evidence card

Keep the claim narrow enough to check. 'Work ruins every week' is too broad. 'Stress was higher in three check-ins that mentioned the Thursday deadline during May' gives you a timeframe, a count, and source language you can revisit.

Then add a counterexample from the same kind of day. A calm Thursday without the deadline matters more than an unrelated Saturday. The purpose is not to defeat the read. It is to show the boundary of what the record actually supports.

Build your evidence card

Work from a single read. Copy its wording before you start so it cannot shift as you gather examples. Define the period it covers and identify the rule for comparable moments, such as the same weekday, the same kind of event, or the same check-in field.

Use source language sparingly. A short phrase and date are usually enough to relocate an entry. If copying a passage would expose somebody else's private details, summarize the relevant context and keep the full source inside your private record.

1. 1

   Freeze the wording

   Copy the exact observation you are checking before gathering evidence or changing its language.

2. 2

   Set the timeframe

   Write the first and last date covered so older or newer entries do not drift into the sample.

3. 3

   Attach support

   Add at least three dated moments from distinct periods and record the specific words or fields that support the observation.

4. 4

   Find a counterexample

   Use the same comparison rule to locate one moment that does not fit the proposed thread.

5. 5

   Mark gaps and decide

   Note missing check-ins or fields, then keep, narrow, reject, or leave the read open.

Watch for overlap being mistaken for cause

A recurring overlap does not show that one thing caused another. Stress and short sleep recorded in the same check-ins can support a same-day observation. It does not automatically support a claim about what happened the following day or why the stress changed.

Keep time direction out unless the product and the record establish it. Words such as caused, led to, triggered, and predicted need more than recurring entries. For a personal evidence card, 'showed up alongside' or 'appeared in the same check-ins' is usually the honest phrasing.

• Name the period instead of implying the pattern has always existed.
• Use the same comparison rule for support and counterexamples.
• Mark missing check-ins rather than treating silence as a good day.
• Keep same-day observations separate from later-day claims.
• Let the person accept, narrow, reject, or reinterpret the read.

Separate product behavior from research support

No research or outcome claim has been approved for public use yet. The Daylogue evidence page separates product verification from bench validation, human testing, and outcome support. That ladder prevents a working feature from being mistaken for proof of a personal benefit.

Your evidence card should do the same. A source moment can support the wording of a personal read. It does not prove that seeing the read will change behavior, improve how someone feels, or produce a reliable forecast. Those are separate questions requiring separate evidence.

Use risk frameworks as questions

The NIST AI Risk Management Framework is intended for voluntary use. It helps organizations think about risks and trustworthiness, but it does not certify an individual journal read. Use its risk mindset to ask how a mistake is found, who could be affected, and how easily the wording can be corrected.

For a personal journal, the safest destination for an evidence card is beside the private record it describes. Keep names and sensitive passages inside the private record. A public evidence card can use neutral handles. Traceability should reduce uncertainty without creating new exposure.