Ahmad Dibo© Notice

B1

01 Context

I wear the same shirt every day, alternating between eleven near-identical black shirts. I have always called this a way to cut the morning decision: if every shirt is the same, choosing between them costs nothing.

Key question

Is there an extractable pattern from which shirt I chose in the morning?

02 Methods

A year of mornings was reconstructed after reporting ninety days of behavioral habit against observed weather. The rest was then generated forward from a model fit to those days. Each morning records what I did (how long I checked the mirror, grooming, whether I trained), what I reported feeling (confidence, dread, energy), and the day itself (weather, schedule, sleep, whether anyone would see me). Shirts are coded best, middle, or worst, and a model tries to recover which tier I picked. Always guessing the most common tier gets 44%.

Selections distributed 42% best, 44% middle, and 14% worst.

  • Fig. 01 Permutation Importance
  • Fig. 02 Learned Decision Tree

03 Findings

Inside the simulation my account does not hold. The choice tracks measures I never reported controlling for.

The worst shirt appears on 83% of mornings meeting two conditions jointly—a day on which no one will see me, and a day on which my headspace is low—against 8% of all other mornings, a gap of 75 percentage points and better than a tenfold concentration.

Mirror-check duration tracks tier more strongly than any other input, at 76 seconds on best-shirt days versus 35 on worst, a difference of 41 seconds, or roughly double. How long I looked is the strongest signal in the dataset.

Given only what I did and none of what I said I felt, the model gets 76% against the 44% baseline. Adding the self-report takes it to 79%, a margin too small to separate the two. What I reported feeling added nothing the behavior had not already carried.

  • Fig. 03 Cross-Validated Accuracy
  • Fig. 04 Recovery by Tier
  • Fig. 05 Hypothesis Summary

04 Discussion

The error is worth stating plainly. I reduced the choices and removed the friction of choosing between them. I removed the deliberation but had left the decision in place.

With eleven near-identical shirts, the decision had nowhere visible to go, so it moved to the mirror. Whatever I settled on in those seconds of looking reached the closet as a tier, and I filed it as habit.

A uniform is adopted to make the morning uninformative. Mine made it legible. Because the shirts barely differ, any consistent preference among them has to be carried by something outside the shirt: the day ahead, who would see me, how I felt about being seen. The closet became an instrument, recording a state I had stopped reporting to myself.

Fewer options relocated the choice to a place I had stopped checking.

05 Live model

Open the model to compose a morning by dragging conditions into the day and watch the learned tree return the predicted tier I would reach for.

  • Live model