Personalization · The signals · Component

The accuracy

How often each inference is right — because a wrong inference is experienced as being misunderstood.

The deliverable

What it is

An attribute inferred at seventy per cent accuracy is wrong for three people in ten. Whether that is acceptable depends entirely on what happens when it is wrong.

The cost of a wrong inference is asymmetric: a missed opportunity is invisible and a wrong assumption about someone is not.

One level in

What it is made of

Each element is a constituent part of the component. Follow one to see the attributes it carries.

  1. The rate

    How often the inference is correct.

    3 attributes: Accuracy · Sample · Measured

    Learn
  2. The cost of wrong

    What happens when it is incorrect.

    3 attributes: Consequence · Severity · Required accuracy

    Learn
  3. The threshold

    How confident the inference has to be before it is used.

    3 attributes: Confidence threshold · Below threshold · Implemented

    Learn

Below the confidence threshold, show the default rather than a guess. A confident wrong guess is worse than neutral.