Lets say for example we set a Target, that the average Blocksize is always 80% of max. Blocksize and adjust the max. Blocksize by max. +-20% all 2016 Blocks to meet this Target.
This would ensure that Blockchain space always remains scarce, therefore ensuring TX fees for fast transaction, by at the same time ensuring that it will always be possible to make a transaction.
I'm looking forward to learn why this wouldn't work.
The problem with this is that some pools still mine blocks where the only transaction is the one that awards them their subsidy, and that can really poison averages.
Than change it to The average of the biggest 50% of all Blocks mined every 2016 Blocks. That would also mean that at least 50% of all miner would have to agree that a increase of the max. blocksize is necessary and also ensuring no minority can keep an increase from happening.
So (numbers changed a little):
Target: Average Blocksize of the biggest 1008 Blocks is always 90% of max. Blocksize
Adjustment: Max. Blocksize, max. +-20% all 2016 Blocks to meet this Target.
Isn't it amazing the difference adding one small caveat makes. This is why I'd default to trusting Gavin's judgement, because everyone is trying to smash all of these ideas in his face and he has to make decisions that protect brialliant ideas from naive attacks.