Some signals deserve more trust

Imagine deciding whether to carry an umbrella.

You look outside. You check the weather app. You ask a friend who is usually wrong about rain.

All three give you information, but you do not trust them equally.

What you see outside might get a high weight (you trust your eyes). The weather app might get a medium weight (you somewhat trust it). Your unreliable friend might get a low weight (she is almost always wrong about the weather).

If your friend is reliably wrong, the connection could even have a negative weight. Their warning would then push the decision away from carrying an umbrella.

That is the idea behind a weight: some inputs should influence a decision more than others.