Democracy for lunch

“The History of every major Galactic Civilization tends to pass through three distinct and recognizable phases, those of Survival, Inquiry and Sophistication, otherwise known as the How, Why, and Where phases. For instance, the first phase is characterized by the question ‘How can we eat?’ the second by the question ‘Why do we eat?’ and the third by the question ‘Where shall we have lunch?”

Douglas Adams, The Restaurant at the End of the Universe

As well as being the mark of a sophisticated civilization, the question “Where shall we have lunch?” has another interesting feature: It’s one of the most common applications of small-scale democratic decision making we regularly run into in our lives.

Suppose we’re a group of friends or coworkers trying to decide on a venue for a group lunch. How do we make this decision? And how should we make this decision?

The most important step in the process is the one everyone is already doing: Talk it out. People can raise options, either well known ones (“Why don’t we go back to our usual Thai place?”), or new ones that might not be broadly known (“There’s a great new steak restaurant that’s just opened up”). Importantly, people can also veto options, e.g. “I have a peanut allergy, so Thai food tends to be an excitingly lethal game for me”, or partial vetoes such as “Only if the steak restaurant has a vegetarian menu”.

Eventually after a number of expansions and contractions of the list of options, you tend to reach an an impasse: There are multiple options that everyone finds acceptable, but no obvious winner.

At this point you could keep talking about it until some sort of consensus is reached, but at this point people are probably bored of talking about it and any remaining disagreement is down to fundamental matters of preference and taste rather than something you can easily talk people around to anyway.

When you reach this state, how do you finally pick a venue?

In my experience usually what ends up happening is that someone (often me) gets fed up and says “OK, I propose we go to this option. Does anyone want to object?” and attempt to build a consensus starting from there. This is OK as a solution, but tends to prioritise the needs of the most assertive and/or least patient members of the group.

Another, slightly less common, solution which is better at respecting everyone’s desires is voting: Everyone in the group who has an opinion on the matter votes on the options and the winning option is chosen.

But how do we vote on these options? Casual use of voting tends to devolve to bad choices of voting systems, usually plurality voting (everyone picks their most preferred option, you go for the one the most people picked). This is far from the best option, even for relatively low importance decisions like lunch.

Instead there are two relatively simple variants on it, either of which are a strict improvement. Which one you should use depends a bit on context, but there’s a very simple test to decide between them: Just answer the question “Do we definitely want to go along with the majority opinion?”

The answer to this question isn’t obviously yes, and depends on a number of factors, the biggest one being how often we are making this sort of decision.

If this is our weekly lunch then we actually don’t want to always follow the majority opinion: If we’ve got 4 vegetarians and 5 meat eaters, going out to a burger place (they do veggie burgers! It’s fine and we’re totally respecting your preferences!) every single week on the basis of a thin majority is a bit harsh. If on the other hand you’re planning the big annual dinner then you probably want to go with the majority view (though ideally you’d have discussed things enough that you’ll find an option with more than a bare majority).

Once you’ve answered that question, I think there are two obvious best options: If you want a majority outcome, use approval voting. If you instead want a fairer spread over many decisions, use random ballot. These aren’t the best options for everything, but for low impact single-winner decisions like this where simplicity of voting system matters a lot they’re very hard to beat.

These work as follows:

Approval voting is normal plurality voting where people can vote for more than one option. You go through each option and everyone who likes that option raises their hand. The one that got the most hands wins (break ties arbitrarily, or by running another vote between the tied options). People can raise their hands for as many different options as they like.

Random ballot using chance to spread out the decision making across the whole group: You pick one of the group by lot and go with their favourite choice. Over multiple decisions this will tend to average out – if 40% of the group are vegetarian, you’ll pick highly vegetarian friendly options about 40% of the time. If you’ve got that one guy in 10 who is completely obsessed with that Brazillian barbeque joint, you’ll go there about 10% of the time.

Random ballot is a little trickier to implement than approval voting, because you need some sort of randomization device, but almost any device will work. The two easiest options are to draw cards from a deck and have high card choose and to put pieces of paper into a hat and pick one out. The latter is better but slightly fiddlier. The former has the problem that by picking the person and then making them choose you put them on the spot a bit and make them feel much more accountable for their choice, which will tend to bias them towards the majority and reduce the proportionality. That may also be a feature for some people.

Either of these will be a significant improvement over normal plurality voting, and neither of them are particularly hard to use. Obviously the random ballot solution is dearer to my heart, but for many groups approval voting will be an easier sell, and for rare events it’s definitely the way to go because you can’t rely on the randomness to average things out.

You can also use these systems mixed in with the initial talking phase rather than keeping them clearly separate – if you want to use approval voting, just let anyone call out a new option, and discuss whether you want to veto it before voting. If you’re using random ballot, use the version where you pick a dictator by lot and then have them lead the discussion and listen to suggestions and vetos but they make the final decision.

An Amusing Variation

There’s an idea called Quadratic Voting which works by letting people buy votes, with the cost being proportional to the square of the number of votes: If one vote costs you £1, then two votes costs you £4, 3 votes cost you £9, etc. The money paid is then distributed amongst the voters equally.

I’m fairly ambivalent about quadratic voting in general, but it’s actually quite well suited for the use case of going for lunch, because there’s an obvious thing to do with the money spent: Put it towards the bill!

You can combine this idea with either of the above voting systems. For manual counting it’s actually slightly more convenient with random ballot, because you don’t have to do arithmetic: Either people get dealt extra cards or get to put multiple items in the hat.

Why am I writing about this?

I’d like to see more casual use of democratic methods, and the use case of lunch is nicely illustrative of how that can come about: Voting doesn’t have to happen at large scales with big life changing (and destroying) results. It can happen at small scales whenever we’re hanging out with our friends and want to make a decision, and it can make life easier when it does.

In general I think we tend to assume this sort of process and formalism is very heavy weight and only appropriate for Serious Decisions, but I think that overstates the difficulty a lot, and this sort of casual use is important if these tools are ever to become widespread.

Also, it really annoys me when we struggle to decide where to go for lunch.

This entry was posted in voting on by .