Insights methodology

The Insights page helps you notice patterns across your journaling — recurring triggers, beliefs you return to, which integration actions seem to coincide with the biggest drop in emotional intensity, and how intensity changes over time. It exists to support self-reflection, not to diagnose or recommend treatment.

The only inputs are the fields you write yourself:

• trigger_tags, trait_tags, action_tags
• core_belief
• intensity_before and intensity_after (0–10)
• created_at (entry timestamp)

No demographic, location, or device data influences the computations. Cross-user data is never used.

All formulas run client-side after your encrypted entries are decrypted in your browser. There is no server-side scoring.

• Top triggers — count of each trigger_tag across all your entries, sorted descending, top 8 shown.
• Core belief heatmap — count of each distinct core_belief string, normalised to a 0–100% bar against the most-frequent belief.
• Trait distribution — count of trait tags grouped by a fixed list of dark vs. golden traits, shown as a percentage of the total.
• Action effectiveness — for entries where both intensity_before and intensity_after exist, we compute the drop (before − after) and, for each action_tagattached to that entry, accumulate the average drop. This is a correlation, not causation: an action may co-occur with a drop without causing it.
• Emotional intensity trend — a chronological line of intensity_before and intensity_after with no smoothing, no forecasting, no thresholding.

• Patterns only become meaningful after roughly 20 entries.
• Tags are user-defined free-text; spelling variations are counted as different tags.
• Action effectiveness reflects co-occurrence, not therapeutic efficacy.
• Insights cannot detect crises. If you are in distress, please contact a qualified professional or your local emergency service.

The EU AI Act (Reg. (EU) 2024/1689) defines an AI system as one that, with varying levels of autonomy, infers outputs from inputs. Onwards Upwards Insights uses fixed arithmetic (count, sum, average) over your own data. There is no model training, no inference, no probabilistic output, no automated decision-making within the meaning of GDPR Art. 22.

If we ever introduce a learned model (for example, AI-assisted summaries), we will publish a separate model card here, re-consult the DPIA, and require fresh, granular consent before your data is processed by it.

• Opt out: turn off Insights from Settings → Insights. With Insights off, no computations are performed and the page displays an opted-out state.
• Delete: erasing an entry removes it from every chart on the next page load.
• Export: download the raw data behind the charts from Settings → Export your data.
• Human review: if you believe a chart is misleading or want a human to explain a result, email insights@onwardsupwards.app. A person will respond within 5 business days.

Changes to the formulas above are tracked in version control and reviewed by the Data Protection Officer before release. The internal change record lives under docs/compliance/. This page is reviewed at least annually or whenever a formula changes — whichever is sooner.

Methodology questions: insights@onwardsupwards.app.
Data-protection questions: dpo@onwardsupwards.app.