Read this first: ICC
Expected Donation Score
In my opinion, the most obvious concern is: if a donor can withhold their donation, what's preventing them from making a very large pledge and not paying after the work is published?
My proposed solution to this is to assign each donor a score that would roughly predict the amount that a given donor will give based on their pledge amount and past donations.
Lets consider an example:
A fraudulent donor made a large pledge for proposal A
that they don't intend to pay.
I, as an artist, am choosing what to work on, and I see that
proposal A has a much higher payout estimate than proposal
B, and work on A as a result.
Once I release my work, I end up not getting paid nearly as much as I
would have for B, which, in this example, has mostly genuine pledges
(though I cannot know that at this point in time).
This is where the "Expected Donor Score" (EDS) comes in.
One potential formula for this metric is $${S_0 = \frac{D}{P}}$$ where ${S_0}$ is the coefficient, ${D}$ is the sum of a donor's past donations, and ${P}$ is the sum of a donor's past pledges. But alas! If the fraudulent funder has previously fulfilled their pledge with a donation of $0.01, then they have a maximum EDS, and we're back to square one.
To fix this, let's scale the EDS by the amount paid: $${S_1 = D\frac{D}{P} = \frac{D^2}{P}}$$ Now, while the fraudulent funder still has an EDS of 1, a genuine funder who has paid $120 of their proposed $200 over the past 2 years has an EDS of 72, greatly outshining the fraudulent funder. 72 is a really big number though, and over the course of many years, it could be reasonable for a third donor to spend thousands of dollars total, making their "influence" many orders of magnitude higher than even the donor who has spent $120.
Let's try adding a time scaling factor. $${S_2 = \sum_i e^{-t_i} \frac{D_i^2}{P_i}}$$ where each payment/offering pair is scaled by how recent it was, ${t_i}$ being the amount of time that has passed since payment ${i}$ in months (an arbitrary time unit). Doing some napkin math, it seems like this helps with balancing the influence of long-term donors vs medium-term donors. We could add a few scaling and offset factors here and there, but for the most part, this seems to solve the problem at hand.
It's worth noting that this is definitely not the only solution to the stated problems, and almost certainly not the best! Perhaps it would be worth allowing artists to mess with this formula in order to let them use whatever they believe best predicts payouts.
One last potential issue I want to address is how fund offering amounts scale the EDS. Someone who has paid $120 over the course of a year suddenly declaring a $2000 fund offering is a bit suspicious, so we should incorporate that into our final calculation somehow. However, I think it's fine to leave the EDS as it is for now, and figure that out later if it actually becomes a problem.
As it stands now, the EDS roughly maps to the amount of money a donor is likely to provide based on all of their past donations.
The calulation of this score is critical to the functionality of converse crowdfunding, as it is what allows ICC to still work even though donors are allowed to retract their entire pledge after the work is published.
TODO
Consider the following: a fraudulent artist makes a proposal that they do not plan on fulfilling, and some donors declare pledges. In this example, the artist fraudulently "publishes" a trivial or low-quality work (say, a blank white image file rather than a page of a comic) just to claim the pledges. How can the system be built to minimally punish these donors? It's easy to say that the donors should have been smarter, and should either pay the amount they offered or retract their entire pledge and just take the hit to their EDS, but ideally, the system should try its best to nullify any incentive that would make the donors want to pay the fraudulent artist. In the long run, this type of fraudulent artist wouldn't last long, but one could easily imagine a washed-up one-hit-wonder artist making one last grand proposal, to which many donors offer pledges, just to run into the above situation where the work is low-quality or nonexistent, intended only to trigger the donation invoice.
Could the EDS be adjusted to mitigate these scenarios? Would it require a totally separate mechanism?
In this case, perhaps we could normalize the offerings? If almost every funder withholds part or all of their compensation, we should be able to use that to quantitatively measure the extent to which the producer "failed" to deliver.
This might be solved by making the EDS unique to each funder-creator relationship, since it's unlikely that the funder cares about their EDS when the creator has tanked trust. However, what about the case where there's plausible deniability, and there might be hope for the creator to make something legitimate in the future?