[anish_m]

I worked on one of these Fraud detection systems at Airbnb. Statistically, You are in 1% population with the your usage patterns. For most part, all ML models which are used to detect fraud have to take a trade off between precision/recall - meaning how often they let good guys suffer vs let bad guys in. BUt what most of these systems are missing is the appeal and escape path.

[spacebanana7]

> But what most of these systems are missing is the appeal and escape path.

Are there any such systems that do have effective escape/appeal paths? If so, what do they look like?

[sneak]

Even if they exist, I am willing to bet they consume the limited time and energy of the people whose only crime was not living the same lifestyle as 99% of everyone else.