Loading Ableway
Routing…
Loading Ableway
A false "accessible" can strand someone for hours. A false "not accessible" costs an afternoon. The asymmetric consensus rules behind every Ableway score.
Greyson Kline · Ableway
Every accessibility score hides a wager. Ours is placed deliberately: the two ways a score can be wrong are not equally bad.
If we say a place is accessible and it isn't, someone may have planned their day around that answer — ridden across town, arrived at a flight of stairs, and gone home. That's not a wrong answer on a screen. That's a person stranded.
If we say a place is not accessible and it actually is, someone loses an afternoon to a detour. Annoying. Fixable. Not the same class of harm.
So we don't score "yes" and "no" symmetrically. We hold "yes" to a higher bar.
When the community reviews a feature — a step-free entrance, an accessible restroom, a ramp — each answer carries weight, and the weighted evidence has to clear a confidence threshold before it becomes part of a score. Those thresholds are different in each direction:
Nothing moves into the score below its bar. And for the features where a wrong "yes" hurts most — entrances, restrooms, elevators, ramps, level boarding — the gating is stricter still.
We ask a third answer most systems skip: I don't know. An "unsure" counts as half a yes — it's weak evidence that the feature is at least visible and not an obvious failure, and it moves the needle without pretending to be certainty.
What it doesn't do is inherit the voter's reputation. A trusted community member's "I don't know" is not more unsure than a new member's. Doubt is doubt. Confidence is what earns weight.
Votes from established community members carry more weight — up to double at the top trust tier, with a bonus for people whose expertise is in the category they're voting on. That weight is guarded on three sides:
Here's the rule I care about most. For any feature without enough evidence, the default assumption is route through with caution — not avoid.
Routing around unverified features is the same class of bias as routing around unverified neighborhoods.
If we treated "no data yet" as "bad," we'd systematically starve exactly the places that haven't been reviewed yet — often the smaller, newer, more local ones. Under-review isn't a verdict. When evidence genuinely conflicts, the dispute stays visible as its own state rather than being averaged away into a misleading middle.
Photo analysis can pre-fill a report — it's good at noticing that a curb ramp appears in a picture. But in our scoring it only ever applies before any human has voted, and it weighs less than a person: a detected "yes" adds a fraction of the confidence a human "yes" would, a detected "no" a smaller fraction still.
The deeper limit is honesty about what a photo can know. An image can show that a ramp is there. It cannot measure whether that ramp meets a slope standard — and we will not pretend otherwise. A camera is a presence detector, not a measurement instrument, and it will never outrank the person who actually rolled up the ramp.
All of this adds up to a language rule we hold ourselves to: Ableway scores are community-reviewed. We don't say "verified accessible," and we definitely don't certify anyone's compliance with anything. What we offer is the next best thing and arguably a more honest one: a growing body of reviews from people who moved through these places in their own bodies, with the confidence of each answer worn on its sleeve.
The thresholds above are live in our scoring engine today, in version 0.7 of the platform. They'll be tuned — thresholds are parameters, not scripture — but the asymmetry itself is a principle, not a parameter.