Ideally. However, even if AFCIs were to be 100% effective*, they would only stop a small percentage of actual dwelling fires. The remaining chunk would still present life and death situations- where as for 2x-3x the cost of AFCIs, we would eliminate near 100% of those life and death situations.
*Ok, preventing fires in the first place, that is a valid argument. However, the bulk of electrical fires come from glowing connections (joule heating)- with the vast majority of them produce no arcing signature. For the sake of the argument even if 100% of them did, there is a huge question if AFCIs will be able to catch them:
https://www.youtube.com/watch?v=-rfqqNg-uVE
Why so? Because AFCIs have to be able to distinguish between current ripples from appliances, motors, switches, light bulbs burning out, power supplies, ect vs an actual loose connection. Wave forums that look almost identical- if not totally identical- must be discriminated by a woeful lack of computing power in residential grade AFCIs. As a result, manufactures must choose between making them more sensitive increasing the odds of tripping on an actual problem
while simultaneously nuisance tripping on a variety of appliances or decreasing sensitivity saving headaches for electricians and occupants while increasing the odds of an actual problem going unnoticed.