The
Michigan/Florida situation is genuinely tricky, both politically and as a matter of moral philosophy:   On the one hand, it is deeply
problematic for the Democratic Party not to count two entire states, whose voters are not fully to blame for the actions of their leaders; on
the other hand, these states did break DNC rules that other states
complied with, that both candidates knew about in advance, and that HRC's people actively supported.   A
rock and a hard place.

Even if one wants to "count" the states, the results of the two unsanctioned contests that occurred in January are clearly not reliable indices of public support for the candidates.  The results are, in fact, highly unreliable, because the candidates didn't campaign in the states and because many voters doubtless relied upon the fact that, at the time the elections were held, the DNC's rules made the elections meaningless.  The results are also particularly unfair to Obama, who was disadvantaged as the challenger by abiding by the mutual agreement not to campaign in the two states.  To use the results of the two unsanctioned contests as if they reflected real elections is not a legitimate means of "enfranchising" the two states -- scarcely better than arbitrarily assigning them random election results would be.

The challenge is to award delegates based upon an estimate, as accurate as possible, of the result that would obtain in a fully contested race between Obama and Clinton in Florida and Michigan.  Here is, in my view, the best and most feasible way to do that, using "baseball-style" arbitration (so named because it's the process for resolving certain salary disputes in Major League Baseball).  It would work like this:

Have each campaign, Obama's and Hillary's submit number reflecting the percentage of votes each believes it would get in each state.  Each campaign could support its claimed breakdown with evidence, such as polling data, results from nearby or demographically similar states, or any other relevant information.  

Then, have a panel of non-aligned judges (say non-endorsing Democratic Senators; neutral political scientists, or the like) evaluate the competing estimates, and select, for each state, the more plausible virtual election result.  Thus, say, if Obama's side predicted a 54-46 win for him in Michigan's and Clinton's side predicted a 55-45 win for her, the arbitration panel would decide which "offer" was more credible, and delegates would be based upon that breakdown.  Having a neutral panel evaluate the bids -- standing ready to pick the offer that is more reasonable -- serves to discipline both sides, and prevent them from making offers that are too-self serving.  Thus, Obama's side wouldn't claim a 70-30 victory in Michigan; or Clinton a 80-20 victory in Florida, because the arbitration panel would reject such extreme submissions in favor of a more likely one from the other side.

The full delegations from the two states would then be seated, with delegates pledged to the respective candidates based upon the percentages selected by the arbitration panel.

Obviously this is a funny way indeed to decide an election.  But it is more fair, more feasible, and less out of sync with actual voter preferences than any other available procedure.