More detailed explanation why we souldn't give "pay back delegates" a perspective right from the start is here https://bitsharestalk.org/index.php?topic=5868.msg78966#msg78966Goals (arranged acc. to subjective importance, most important first):
(1) Minimizing the influence of (obviously) bad actors.
(2) Prevent centralization. Source for centralization
https://bitsharestalk.org/index.php?topic=5775.0 There might be a free rider (group trap) problem depending on how high the tx fees are that you get back for voting for the giving back delegate. I think it is likely that people can not estimate how much they can loose (together with everyone else) in relation to what they gain and simply take the additional gain because it is the most obvious benefit. Then the free rider problem is there no matter what the estimate is approximately in reality. (2) is interrelated to (1) as a bad actor could use the pay bcak to voters scheme as a means to gain control. But decentralization also has benefits of it's own so it is an own own point.
(3) Maximum representation (= percentage of shares equals percentage of approved delegates).
Instruments / suggestions:
(a) delegation of votes (= your voting power is distributed among those you vote for)
(b) approval voting (= you can give as many delegates as you want (variant: only a limited amount of delegates) your approval, each approval vote has the same weight.
(c) negative votes (= vote against a delegates in style (a) or in style (b)). Variant (d): Negatives votes could be weighted more than the positive votes (e.g. 1 negative vote + 2 positive votes = 0). Variant (e): limit the amounts of negative votes in case negative votes are style (b) / via disapproval voting.
Effects/propositions:
- (a) and (b) equally lead to (2) if people perception is that they gain more from voting for the delegates that pays out the most tx fees to his voters compared to his competing delegates and/or if the actually gain/looses estimate is in favor of such behavior.
- approval voting is better to minimize the influence of a potentially bad actor.
- In the end negative votes is the same as positive votes. Just the procedure is different. Right? (didn't think long about this)
I don't know if "the give back to voters" attack can be prevented without a "social ban". I don't know a solution. Ask a game theory specialist...