First of all, a 'Guid New Year' to you.
This post is a little more technical than most. As always, the devil is in the detail and we need to get into the detail to understand the issues. I hope you will plough through to the end.
If you've just watched a movie (and who hasn't over the Christmas period?) then you can say quite a lot about what appeals to you and what doesn't. Is the story engaging or trite; are the characters believable or implausible; are the stars attractive? What about the production values, or the score? You can see that there are many dimensions on which you could rate any movie. And so it is with political opinion.
Basics
First of all, a reminder about how the panocracy voting system is intended to work.
Let's start with a simplified example. Here we have a panocracy of 5 voters and a proposal they are voting on (see diagram 1 below). Each voter's political opinions are positioned at a point in a 'political space'. In this simplified example, the space has just two dimensions. The proposal (RFC) they're voting on is also positioned in the same space and each voter's distance from it is calculated. If the voter is 'close' to the proposal then their vote is counted as 'for'; otherwise it's counted as 'against'.
There are a few(!) practicalities to be examined which we deal with now.
Abstentions
Can we represent abstentions?
If we expand the circle that represents the 'cut' into an annulus or ring (diagram 2)
then an opinion that falls within the annulus can be interpreted as 'neither for nor against'. The inner and outer radii of the annulus will have to be set consistently. At the moment it seems sensible to set annular width to be proportional to the (normalised) variance. This would result in a relatively constant fraction of abstentions or 'undecideds' on any vote.
Dimensions
For simplicity in the diagram above we have shown just two dimensions or parameters.
In real life, political opinion has many parameters as in our movie example.
The voting system must avoid degeneracy (in the mathematical sense). For example, two voters' views may be identical on economics, etc. and may differ only by their opinions on immigration. If a proposal relating to immigration arises then the system needs that dimension so as to distinguish between those voters' opinions on the matter.
A large number of dimensions will likely required to do this in the general case. It may also be necessary to add new dimensions as experience develops.
Normalisation
The axis scales must be normalised, which is to say scaled so that all the dimensions have equal weighting.
Under the assumption of a normal distribution we could calculate the inner radius of the ring to be such that at least half the population would approve a proposal that coincides with the mean.
Independence
In the above graph independence is represented by the axes (basis vectors) being orthogonal (at right angles): if a voter's opinion moves to the right or left on axis 1 then that has no effect on his axis 2 position. This, together with normalisation would mean the 'cut' ring is circular and keeps the voting process simple!
Independence is however difficult to achieve just by making an arbitrary choice of which characteristics each axis represents. We need to look at correlations between axis data and use those to apply corrections. For example, we might observe that there's a correlation between the opinions on an urban/rural axis and a religious axis: those with more rural preferences might also tend towards religious authority and vice versa.
As it happens, this is a common problem in data analysis especially in the context of machine learning and there are well-established ways of dealing with it as the next section shows.
Feature Extraction
The way this is done is by using Principal Component Analysis (PCA) or here for a more mathematical treatment.
To summarise, we start off with an arbitrary set of (say) 20 parameters to characterise each voter’s opinion and then use PCA to encapsulate most of the original information but which are now independent of each other and normalised so we can apply the simple vote-counting technique shown earlier.
This makes vote counting consistent.
In the case of our panocracy, each voter's opinion will be captured via many parameters (see below). So Joe Soap's opinion along the axis of 'Conflict versus Negotiation' might be near average but his opinion on 'Religionism versus Secularism' might be a long way from the average, and so on for many other parameters. Tens or hundreds of parameters may be required to properly characterise someone's political opinion but the PCA procedure will ensure every opinion on every issue is counted equitably.
Because the dimensions arrived at by the PCA process are not the same as the originals, publicising the positions of voters along each axis will not compromise voters' privacy or confidentiality. In other words, a voter's ID (if they even have one) won't be included in published data and it won't be possible to 'reverse engineer' it if redundant information on voters' opinion is discarded.
In view of the ever-changing nature of the realities of the world - for example we didn't have a fast internet and social media even as recently as a quarter of a century ago - the characteristics that allow panocracy to define voter preferences are likely to change with time.
So PCA would be applied continually to voter opinions as a basic function of the panocracy.
Ultimately, this system will judged on whether it votes the way every voter expects and any voter can check this at any time.
Assumptions and Problems
In the foregoing we've assumed that opinion along each axis is unimodal. This may not be true, for example the case of 'polarised' views is likely to be bi-modal. This may or may not be a problem in practice. It may well be possible to identify multimodal distributions and split them into multiple unimodal dimensions.
However, PCA may not itself spot obvious clusters - where opinions happened to be distributed in an unusual way (see Principal component analysis by Lever, J., Krzywinski, M. & Altman, N. Principal component analysis. Nat Methods 14, 641–642 (2017).
Let's say there's a proposal to introduce (or rescind) the death penalty for murder. Note that we're not here to argue for or against this proposal, merely to look at how it might be handled in our panocracy. (Even in the United States, public and judicial opinion on this varies with time and location. The death penalty was removed and then reinstated in the 1970s.)
At first sight, this is a binary choice so we might expect the population to be divided into two camps each with about half the population. Is that assumption justified?
In our panocracy and before the proposal ever reaches a vote, there's a long debate at the RFC stage about costs (unsurprisingly, capital cases are more expensive to prosecute), proportionality, checks and balances to reduce miscarriages of justice (which are not reversible!), the effects of 'closure' (or revenge if you prefer) and so on.
So perhaps people are not so polarised as we might have expected – the lower costs of non-capital trials might sway those who want to reduce public spending. Of course, the cost of long term incarceration might outweigh this!
Positioning a Proposal
A significant unknown in all this is how to position a proposal (i.e. RFC) in the political space. Clearly, it's critically important to position a proposal 'correctly' on each axis – in other words so that its position accurately reflects its political status.
We could position it on each axis according to its content. For example, a proposal (RFC) to send military aid to the Ukrainian conflict would be expected to appeal to those of an interventionist bent, but perhaps not to those interested in foreign trade (with Ukraine or Russia). It might have religious and moral dimensions too.
There are other ways to determine where a given proposal should sit in relation to each of the axes.
We could take the political position of the proposal's champion or some kind of average of its supporting group and position the proposal there. This is nearer to what we do at the moment with representation where an individual with strong political views votes on our behalf.
Panocracy would be a huge improvement here because of its sensitivity to the issue at hand – the political position of the proposal – and not the hidden hand of party politics.
Implications
It's possible to evaluate a new proposal against what is known about voting intentions – assuming the distributions are made public. This can be done at any time to see if the proposal has a chance of passing.
If the distribution of voting intentions is dynamic – and it should be given that people's opinions and their demography change – then a referendum is merely a snapshot of the prevailing view. A proposal that passed on Monday might have been rejected on Tuesday had it been put then. This is of course true of all legislation however it's determined.
So given that we’d like our laws to be acceptable for an extended period (1 year, 5 years, 10 years, …), a proposal could be evaluated in such a way that it will still pass under reasonable demographic and opinion changes in the specified time frame.
The Voting Parameters (Axes) – A First Cut
In the following sections we’ve tried to lay out a list of the kind of parameters that might be important in voting. Our objective is to specify sufficient dimensions to completely cover the space of political opinion (easy, eh!)
In 1950 L W Ferguson came up with 3 'dimensions' to characterise political opinion: religionism, humanitarianism and nationalism.
Religionism: belief in God and negative attitudes toward evolution and birth control. A voter at one extreme would be a creationist and bible literalist.
Humanitarianism: war, capital punishment and harsh treatment of criminals. At the opposite extreme a voter would strongly oppose all these.
Nationalism: censorship, law, patriotism and communism.
His dimensions were soon complemented and expanded on by Hans Eysenck and many others.
We've laid out some of them here (in no particular order) as a starting point for further work and discussion.
Totalitarianism versus Libertarianism: centralised control of thought and action against complete freedom of these.
Conflict versus Negotiation: this is how a voter would want his panocracy to respond to external aggression – fight or talk.
Religionism versus Secularism: a measure of religious influence in affairs of state – at one end are those who think Church or religious ideology and the State should be the same (e.g. Islamists)
Urban versus Rural: a city dweller will have a different perspective on a new housing estate from someone whose income is from the land.
Interventionism versus Non-Interventionism: the state should exert power outside its jurisdiction to implement its policy or it should keep to its own affairs.
Multilateralism versus Unilateralism: the panocracy should coordinate its external policies with other states or press on independently.
Political force versus Pacifism: at one extreme, force (including violent force) is a legitimate or necessary means of political expression.
Personal self-sufficiency versus dependence on the community: this characterises a voter's position on say, how much public subsidy there would be for home improvement.
International free trade versus Protectionism: for example, zero import Duty versus high import tariffs to protect jobs.
Domestic trade freedom versus Trade equity: on the one hand, businesses should be able trade without regulations, and on the other all aspects of trade should be completely regulated.
Multiculturalism versus Traditionalism: the panocracy should encourage a diversity of cultural ideas or it should preserve and encourage the existing culture.
Human Rights: on the one hand human rights are immutable and fundamental and on the other any claim of human rights may be ignored.
Positive versus Negative Rights: each citizen has rights to prohibit interference by others as opposed to having rights to impose obligations on others.
Rapid versus Slow Political Change: would the voter accept political revolution or at the other extreme no change at all in the political system. For example, replacing the panocracy with an absolute monarchy.
International Federalism versus Separatism: for example as applied to membership of the EU or other umbrella jurisdictions like the World Health Organisation.
Domestic Centralism versus Regionalism: important in societies where strong regional or ethnic identities are political issues.
Support versus Opposition to Private Property.
Labourer versus credentialed elite: the voter may be an unskilled labourer with no academic qualifications as opposed to a highly educated and privileged aristocrat. A labourer would be expected to support a proposal that favours working class interests over those of a banker.
Upper versus Lower Social class: down and outs at one end; royalty at the other!
Religion versus Science: at one end the voter would completely accept 'revealed truth' and at the other they would support the scientific 'philosophy', for example hypothesis testing with real-world data and scepticism of authority, etc.
Focus on Outcomes versus Opportunities: The voter prefers to make adjustments to obtain some predetermined distribution of outcomes (usually equality) rather than to support a meritocratic approach in which outcomes are determined by other factors, e.g. hard work, enterprise or chance.
As a preliminary example, here's a list of characteristics that might matter to a modern Conservative (party) voter as revealed in an unherd comment:
The Nation State, democracy and sovereignty.
Belief in and support for private enterprise and wealth creation.
Taxation that incentivizes risk and entrepreneurship not Brownite redistribution.
A small neutral civil service which goes to work.
Control of our borders and controls on migration to protect broken public services.
Anti welfarism and the spread of benefits in a culture of entitlement.
Strong defence forces.
Resistance to the new progressive ideologies of DEI identitarianism and fanatic Net Zero climate alarmism.
Determination to reform the NHS and the Equality Laws which are warping an imposed multiculturalism.
We can see how each of these fits in with one or more of the above 'political dimensions'. For example, for The Nation State, democracy and sovereignty the voter would be positioned towards the separatist end of the International Federalism versus Separatism dimension.
There are probably many more political dimensions that would further help in identifying the views and opinions of voters.
Caveats
I have to say that I'm not that happy with this list myself! In the course of thinking about the list and consulting various sources, it emerged that the objectives of the political theorists and psychologists whose work it's drawn from are a little different to ours. So we expect this list to undergo considerable change over time, especially as we get deeper into the process of capturing voter opinion.
A major consideration for us is how in practice we're going to capture the opinions of voters in our panocracy. We could just ask questions like 'Do you think human rights are important?' or 'How much does your religious belief affect your life?' but the answers are unlikely to be accurate. Opinion pollsters know this only too well. Asking the right question is crucial to getting the poll results that the client wants.
Some kind of indirect assessment of voter opinion is probably required. If you're unhappy about that, remember that Facebook and others already do this to you. They just don't make a big noise about it.
As always, please feel free to comment on or ask questions about this post. All welcome!