✦ ADHDorNot ✦
a quiet reflection
A non-diagnostic tool that asks adaptive questions and shows you a transparent picture, across the four areas clinicians look at.
How does it work?
This is not a diagnosis, ADHDorNot is a non-diagnostic self-reflection tool. It asks a series of adaptive questions and shows you a transparent, uncalibrated picture of how you answered, across the four areas clinicians look at. It has not been clinically validated, and no website can diagnose you.
Your answers update four connected probability estimates, one each for inattention, hyperactivity/impulsivity, childhood onset, and functional impact, using straightforward, transparent maths (Bayesian updating with published base rates as a deliberately weak starting point). The next question is chosen to be the most informative one given what we already know.
Alongside this reflection, your answers are scored against four published, validated screening instruments (ASRS v1.1, ASRS-5, WURS-25, MEWS) using each instrument's own published cutoff, plus a fifth instrument — the Marlowe–Crowne Social Desirability Scale (SDS) — that detects whether answers may be skewed by a desire to present yourself favourably. This way you can see exactly how your responses map onto tools clinicians actually use, with an honesty check on top. Every result links back to the academic paper it came from.
Read the full methodology & sources →…
✦ ADHDorNot ✦ — Reflection Report
Validated screen concordance
Each row scores your answers against a published instrument using its own cutoff. Green = below threshold; amber = at or above screening threshold.
What the traits mean
Trust indicators
Next steps
Appendix: Item-Level Detail
Each question you answered, which trait it affected, and which screening instruments it fed.
| Item | Question | Your answer | Trait | Screens fed | Direction |
|---|
How ADHDorNot works
A transparent, non-diagnostic reflection tool built on two independent layers: a Bayesian trait estimator and scored alignment with four validated screening instruments, plus a social desirability check.
Layer 1, Bayesian trait reflection
Your answers update four probability estimates, one for each of the four domains clinicians assess. The starting point for every estimate is the published adult ADHD base rate of ~4.4% (Fayyad et al., 2007), but only as a deliberately weak anchor; your own answers quickly outweigh it. Because these domains tend to overlap in real life, the tool shares a small fraction of each answer's evidence across related domains. The next question is chosen to be the most informative one given what we already know.
- Domain probabilities are modelled as Beta distributions, transparent, closed-form, auditable.
- Question selection uses expected KL divergence, picks the option with the highest information gain.
- The result is a smooth, adaptive picture, not a rigid cutoff.
- Its exact thresholds are the author's judgement, not a validated cutoff. They exist only to give a smooth, directional picture. This is why it is paired with Layer 2.
Layer 2, Validated screen concordance
Your answers are also scored against each instrument's own published scoring rule and cutoff, using the full set of items each scale needs (the adaptive selector never skips items required for validated scoring). This is the layer with real clinical weight: every result is anchored to a peer-reviewed paper and a specific validation population.
Built-in honesty checks
Three bias detectors run alongside the main engine and flag concerns in your report when they fire:
Social desirability
Marlowe–Crowne SDS items detect when answers may be skewed by a desire to present yourself favourably.
Inconsistency
Repeated or near-repeated items check whether your answers are internally consistent, not random or careless.
Acquiescence
Tracks whether you tend to agree regardless of content, which can inflate scores on symptom-count instruments.