Liquid democracy and conditional delegation

In liquid democracy voters can delegate their votes, hence its hybrid direct/representative nature. But what if vote delegation as described is not flexible enough? What if voter A wants to delegate to different voters, say B or C, depending on the subject that is being voted on? Vote delegation that is dependent on the vote’s subject matter category is what I call conditional delegation. An example would be

Voter A delegates to voter B if the vote is about foreign policy

Voter A delegates to voter C if the vote is about economic policy

Here foreign policy and economic policy are vote categories. Thus, voters can express their preferences with greater flexibility and precision than with a single, unconditional delegation. It remains to be seen how useful this mechanism is in practice. Perhaps the added flexibility is overkill, complicating things more than the benefits it offers. Nonetheless it can always be offered as an optional feature that more demanding users can exploit.

Throwing conditional delegation into the mix brings with it its own challenges. Here are a few

How is the vote category determined? What vote categorization mechanism can be used such that it does not present an opportunity to manipulate results?

The most obvious method is to include a global service that categorizes votes as they come up, and posts the category for each. This service can be operated by the administrators of the liquid democracy system, analyzing the subject of each vote and coming up with a suitable category. The tally algorithm takes that category as input when resolving conditional delegated votes.  Pretty straightforward.

The problem with this system is what the second question above addresses. Such a global service puts too much power in the hands of administrators, as vote categorization has a direct bearing on tally results via conditional delegation resolution. Even if said administrators were honest, the potential for manipulation would exist, making the whole set up suspect.

A more robust solution is to allow anyone to create a categorization service, and to require voters to choose which categorization service they wish to subscribe to. For convenience, there could be a default categorization service operated by the administrators or by a trusted institution. By opening this service up to anyone that wishes to offer it, categorization is decentralized, anybody can choose which service to use and change it on demand. Thus the potential for manipulation is almost zero, especially if no default service exists.

Taking the above idea even further, one could place the responsibility of categorization on the delegates themselves. In this way a delegate also plays the role of categorization service, and the choice of delegate serves the double role of vote delegation and category service selection. This offers the most distributed and manipulation free option, but it places a larger burden on delegates, as they have to not only vote, but choose a category for each vote as well (although transitivity could also apply to vote categorization).

How does conditional delegation affect loop detection?

As we discussed previously, vote-time loop detection is necessary to avoid lost votes. It works by detecting loops on demand when a voter chooses a delegate, and asks the user to change their selection. But is vote-time loop detection possible with conditional delegation? The answer is probably not.

Vote categorization is not known when a voter makes their conditional delegation, hence checking for loops is uncertain. In theory it would be possible to check loops given all possible categorizations, but that would not make much sense, as the possibility of a loop is not the same as a certainty. It seems undesirable to ask users to change their delegation because of a possibility that could never materialize. So the short answer is that conditional delegation precludes vote-time loop detection, and therefore opens up the possibility of lost votes.

Tally-time loop detection is unaffected of course, the tally will complete correctly as usual.

Should vote categories be fixed and unified? Should they overlap or be mutually exclusive?

The most simple scenario uses a fixed set of categories that are mutually exclusive. Using a fixed set of categories makes it easier for category services and conditional delegations to match. I do not see any reason why a sufficiently rich set of them could not be arrived at ahead of time in a non controversial way.

The question of overlap is a bit trickier. Votes usally fit more than one category, and choosing categories to avoid this will be unnatural. Overlapping categories, although more natural, pose technical problems. Conditional delegation resolution with overlapping categories could yield ambiguous results, if more than one target delegate satisfies its corresponding condition. Reusing the example above

Voter A delegates to voter B if the vote is about foreign policy

Voter A delegates to voter C if the vote is about economic policy

If a certain vote were deemed to be both about foreign policy and economic policy, delegation resolution would be ambiguous.

Recap

We can list the set of technical problems associated with conditional delegation as follows

– ambiguous delegation due to overlapping categories

– lost votes due to

– difficult/impossible vote-time loop detection

– vote subject matter not categorized by delegate/service

– vote not cast by delegate (this is shared with unconditional delegation)

In the next post I will discuss a conditional delegation scheme that can partially address these problems. However, as I’ve stated above, it remains to be seen how useful this mechanism is in practice. Perhaps the added flexibility is overkill, complicating things more than the benefits it offers.


A question I’ve left out is

What logic should be used to express conditions?

I’m pretty sure that anything beyond a specific category (i.e. category => delegate) is overkill. But if you really want to get fancy.. you could use propositional logic/boolean algebra.

2 thoughts on “Liquid democracy and conditional delegation

  1. Someday we might laugh:

    “640K ought to be enough for anybody” Bill Gates
    “I’m pretty sure that anything beyond a specific category is overkill.” David Ruescas

    Anyway, I agree that it seems excessive right now.

Leave a Reply

Your email address will not be published. Required fields are marked *