whysice

A small forensic tool for the moments between click and commit.

The question

When a user says yes, does the cursor agree? Most UX research treats decision as a binary event, click or no click. But the centimeters traveled before that click contain a second signal. The question was whether that signal could be read.

The hypothesis

Decision friction is visible in cursor motion before it is visible in self-report. If hesitation, trajectory deviation, and direction reversals can be measured per-question, then conflict can be located in a flow without ever asking the user whether they were conflicted.

The instrument

Pulse is a three-mode web app. In Design, the researcher builds a flow as a sequence of decision rooms, each a single question rendered as a binary, multi-choice, slider, spread, or stacked layout. In Test, a participant moves through the rooms while Pulse records cursor trajectory. In Report, the session is summarized as total time, average friction, x-flips total, and a count of high-conflict rooms, with per-room analysis and a plain-language insight at the top. The tool is free and runs in the browser at ux-research-pulse.netlify.app.

The method

Friction is not a single metric. It is a weighted composite of six peer-reviewed measures, computed per room:

  1. MAD, mean absolute deviation from the ideal cursor path, weighted at 25 percent.
  2. AUC, total area under the deviation curve, weighted at 20 percent.
  3. X-flips, direction reversals along the horizontal axis, weighted at 20 percent.
  4. Sample entropy, a measure of trajectory complexity, weighted at 15 percent.
  5. Velocity changes, speed instability across the motion, weighted at 10 percent.
  6. Pause count, hesitation moments above a threshold, weighted at 10 percent.

The weights reflect the consensus in the cursor-tracking literature that spatial metrics are more diagnostic of decision conflict than temporal ones (Freeman, 2018; Wulff et al., 2019). Composite scores are banded into four states: calm (0 to 20 percent), uncertain (20 to 45 percent), conflicted (45 to 70 percent), distressed (above 70 percent). The bands are labels, not diagnoses.

What it caught

Illustrative until a real session observation is substituted. In an early run of a five-room flow that included questions weighted toward memory and irreversibility, the instrument flagged a room that the self-report rated “easy” as the room with the highest x-flip count in the session. The participant described the choice as comfortable. The cursor disagreed. That gap, between what was said and what was moved, is the gap the instrument exists to make visible.

Findings

01. Spatial signals lead temporal ones. Across the test sessions observed so far, x-flips and MAD diverge from baseline earlier than pause count or total time. In other words, the hand hesitates before the clock notices. May 24, 2026.

02. Self-report and trajectory disagree most where the question implies loss. Rooms framed around “you can keep one” or “what is heavier” produced the largest gaps between participants’ stated ease and the cursor’s friction score. The instrument is most useful, so far, in flows where the design is asking the user to give something up. May 24, 2026.