Explore authored keyword rules and a conceptual network. Scores do not assess mental health, abusive behavior, intent or cultural traits.
What can one sentence tell you about someone’s motives? This tool checks substrings and draws a fixed four-node network. It cannot establish a diagnosis, prove an attribution error, decide whether a relationship is abusive, or measure a culture. Separate observable behavior from hypotheses about intent, then ask what evidence could change each hypothesis.
At severity 70, “They always asked me to budget” gets base 6 + 1 for “always” + 0.8 for “asked” or “budget” + 70/50 = 9.2. Each keyword group adds once, even if several words match. “They are not toxic” still triggers the label group: negation is ignored. Both scores are authored rules, not empirical findings.
Custom analysis caps 6 + group bonuses + severity/50 at 9.9. Moving any slider instead uses the selected preset baseline (8.4, 6.8 or 9.1; custom uses 7) + (severity−50)/25, clipped to 1–9.9. Thus custom severity 100 gives 9.0 after a slider move, but the worked sentence gives 9.8 when analyzed. Presets start at fixed scores. Amplification and arousal do not contribute directly to the score; their input event can nevertheless replace it.
The graph uses average tension = (severity+amplification+arousal)/3. This changes the frame-node radius to 28+tension/10. Its four nodes and three edges are predetermined; layouts and widths are not evidence of causal influence. JSON records scenario state; PNG captures the graph with a scope label, not the entered claim or full rule explanation. Keep this explanation with any shared image.
Do not use a low score or a benign alternative explanation to dismiss reported harm. CDC includes psychological aggression and control within intimate-partner violence; that definition concerns romantic partners, not every family disagreement. NIMH advises discussing specific symptoms and concerns with a health care provider. Neither source validates this tool’s scores or presets.