This past year I’ve had the pleasure and privilege to contribute to yunity.org: a self-organizing, unincorporated and nomadic team spreading the foodsharing.de movement world-wide.
A key factor in the success of the project so far has been the adoption of Systemic Consensus for group decisions or where autonomous decisions met resistance (think doocracy, iwilldoitocracy and advice process). In this framework there is a real effort to reach consent decisions, but always progress (even if that is an active decision to keep things the way they are).
But this article is not mainly about Systemic Consensus even though I would currently recommend it as the best option for self-organized groups. The following text examines group decision-making frameworks with an informational emphasis (i.e. excluding Autonomous, Authoritarian and chance-based frameworks). By pointing out the main weaknesses of one system, the next is justified. Starting with plurality voting, ending up at Systemic Consensus then speculating even further… read on if you dare, it’s quite cold and has a fair amount of assumed knowledge.
Assumption: The more information we have, the closer we can get to what is real — and that’s a good thing.
Zero-sum majority
Majority voting systems such as plurality voting (“chose one”) and cumulative voting (“place 5 votes”) suffer from multiple negative effects. One is the effect of wasted votes: votes not cast for the winning option confer negligible decision-making power. Another is the spoiler effect: options that are similar lead to the splitting of the vote, ultimately favoring the selection of more unique options. Both of the previous effects lead to tactical voting where voters dishonestly express their opinion, typically to prevent another popular but personally unfavorable option from being selected.
Ranked voting (“rank options from 1 to 5”) attempts to prevent the previously mentioned side-effects but still suffers from tactical voting, surprising complexity and other strange phenomena.
Perhaps the fundamental reason the above systems suffer is one of information loss: in these systems the expression of each voter is a zero-sum game. The support a voter gives to one option detracts from the support they can provide another. They are forced to chose instead of honestly expressing themselves — these systems are at best an attempt to aggregate many individual decisions into one, instead of identifying the emergent group decision.
Non-zero-sum majority
Approval voting (“chose as many as you like”) does not suffer from zero-sum effects as voters are able to express themselves equally across all options. This eliminates the effects of wasted votes, the spoiler effect and tactical voting is reduced. However voters cannot express their intensity of support for each option, only yes/no — again, information is lost.
Range voting (“rate how much you like each”) is similarly free from zero-sum effects and additionally allows voters to express the intensity of their support. This reduces tactical voting even further and permits the most expressiveness of all the majority voting systems:
But there is a problem with majority…
Consent* and consensus
All the above systems are majority focused, asking voters to express what they want. Methods for consent decision making and consensus decision making posit that resistance is more important than support. Not allowing participants to express their resistance and making decisions based simply on support leads to the dragging of resistant individuals and creates unnecessary majority vs minority scenarios where mutually agreeable options exist.
Consent ensures no participant need endure anything they can’t agree to with the inclusion of veto. Consensus attempts to reach a consent as best possible, but failing that it will proceed with the least resisted option.
In consent, the likelihood of veto increases with the number of participants, virtually guaranteeing decision making stasis (i.e. maintenance of the status quo) after a certain point. Whilst consensus is free from the scale limitation that veto brings, both consent and consensus methods are typically conversation-based and eschew voting systems — in this way both are greatly inhibited from scaling.
Systemic Consensus
Systemic Consensus seeks consent and utilizes range voting with a Resistance Rating (e.g. 0 to 10 resistance). Instead of participants expressing support, they express resistance and the option with least resistance is selected. The inclusion of control options alongside proposals in every ballot ensure there is an option for everyone (the Passive Solution being similar to the effect of veto). If the rating leads to two or more options with equally low resistance, those options are re-rated using a scale of support (e.g. 0 to 10 support). However, this author posits that a loss of information is still present within this system as it can often appear that there is no support for the chosen option!
Full spectrum rating, consent-orientation
The inability to express support may lead to artificial relativism: In a range where only resistance can be expressed, participants will be inclined to rate their most favored option with no resistance, their least favored option with ‘x’ resistance and the remaining options somewhere between those options. This can lead to the situation where many options which are actively supported appear to be resisted. The information loss here is exactly opposite to majority voting: where majority voting leads to loss of resistance information and creates false support information, rating just on resistance may lead to loss of support information and create false resistance information.
where majority voting leads to loss of resistance information and creates false support information, rating just on resistance may lead to loss of support information and create false resistance information.
This author hypothesizes that a scale which allows participants to actively express support and resistance (e.g. -5 to +5)** would exhibit a similar rating distribution to a scale which simply allows the active expression of resistance (e.g. 0 to 10 resistance).
In a hypothetical decision making framework utilizing this full spectrum scale, this author advocates the preservation of the Systemic Consensus decision rule: that the least resisted option is always selected, unless there are tied options in which case tied options re-rated on support.
For example, individuals expressing support ratings (e.g. -3)** would be considered as having no resistance (i.e. 0) during the first count where all options are simply assessed on resistance with the least resisted option selected. If two or more options register equally low resistance, all the ratings expressing support but masked as ‘0’ would be recounted at their original value and the option with most supported option selected.
May we fear not the discomfort and complexity from being able to see information — if it is real, let us be brave and face it.
Thanks for reading, goodnight!
*Consent defined here as defined here as the absence of disagreement.
**Use of negative value for support and positive value resistance comes from the convention of describing energy in Chemistry. May help to think about altitudes: skydiving is spontaneous and is a negative altitude differential.
