In which I try to numerically quantify which party I'm voting for and why. Based on a discussion yesterday, it seems to me that voting preference is not dissimilar to the Drake Equation : you have in mind (at least I do) many different criteria which you don't add when considering who to vote for, you multiply them. This means that you can agree 100% with a party based on their policies, but if your trust in them is zero then you're not going to vote for them. It's quite a useful exercise, I think, to try and quantify how much you agree with different aspects of each party - it should at least try and get you to analyse your own choice, if not actually make one initially.
Key points :
- This is a self-analysis. You can use it to analyse your own choices, explain them to others, and set out which areas you need convincing in to change your mind. But you can't claim that these values are objective (except for policies) so it's of limited use in making an initial decision.
- I assume that all parameters deserve equal weight, which is too simple, but what to do when someone thinks a parameter is irrelevant I'm not sure.
- Tactical voting is very difficult for account for in this system.
- You should try as much as possible to independently evaluate each parameter, otherwise you won't get anything useful from this at all.
- This is just an off-the-cuff idea I had that I thought would be worth exploring, and might be totally pointless or wrong-headed.
https://astrorhysy.blogspot.com/2017/05/the-political-drake-equation.html
So this is the Taylor expansion of the Drake Equation!
ReplyDeleteRhys Taylor
ReplyDeleteHi Dr Rhysy, what about the qualities of the individual candidates standing for election in your constituency ?
Adam Synergy Good question. You could apply the equation separately to the constituency candidates as well as the political parties. If the candidate for your preferred party rates highly, then all is well. If they don't, then you get maybe an extra "jerk" factor. You could then just apply this as an additional parameter in the main equation.
ReplyDeleteBut if and how you apply this depends on your weighting and tactical considerations - maybe you don't care that they're a jerk and only care about which party is in power, or maybe it makes all the difference in the world to you. Same for the other parameters - the weighting scheme needs development.
Adam Synergy In a limited sense, any political party is the vector sum of its candidates' voting records and stated policies, which the Taylor Corollary to Drake's seem to encompass fairly well, producing Trust. For Trust is earned on the basis of promises kept.
ReplyDeleteConsider Rhys' column for Labour: the Policies cell is higher than those of the LibDems and Greens on the same row. An enquiring voter would mark up his individual candidate accordingly.
If you take log of individual scores, you are back in additive country. Not sure if that helps in decision making, but adding is, well, more linear than multiplication.
ReplyDeleteI think I've come up with a reasonable way of weighting the different parameters, e.g. policies are generally thought of as more important than behaviour, though not always. For each parameter the user could assign an importance value (the same for all parties otherwise it becomes excessive) on a scale of 0-10, 0 meaning either not at all relevant or that the user can't estimate the value and 10 being of maximum importance. Then the parameters are modified by the importance values (by a simple linear interpolation) so as to be equal to their original input value if the importance is 10 but only equal to 1 if the importance is 0.
ReplyDeleteE.g, if a policy value of 5.6 is specified, and the importance is 10, then P = 5.6 in the equation. But if the importance is 0 then P becomes 1, so it has no effect on the other parameters.
This means that the maximum possible value of V will change - theoretically down to 1 if the user sets the importance of all parameters to 0, but if they do that then there's no point them taking the test anyway. Otherwise, the value of V can be expressed as a percentage rather than an absolute value. I think the percentages should be the same if the user keeps the importance the same value for each parameter (e.g. they might leave them all at the default value of 5, say).
Anyone know a good way to build an interactive web form that can deliver anonymous data ?
Interesting proposition. Allow me to posit a set of slider controls, values 1 to 100. Seven slider bars, all at max values, for a sum of 700, each value is equal to the others, each is thus weighted at 14.28 . Sum of all slider bars gives me the denominator, still have to keep the 1 min value on each slider to make the division work.
ReplyDeleteI can certainly build your form and the back end. I have a seminar to attend, this evening and all of tomorrow, but I'll do some scribbling in the interval.
Dan Weese It would be very cool indeed if you want to rig something up ! No rush, it's not like the political crises are going anywhere... :)
ReplyDeleteSliders are exactly what I had in mind. But I'd stick 'em all in the middle by default, so people feel like they get to increase the importance as well as decrease it (otherwise if they feel that one aspect should be dominant, they have to decrease all the rest, which is irritating and non-intuitive).
What I was thinking would be each slider setting a weight value W (call it "importance" to the users) from 0-10 that modifies each parameter value for use in the main equation. Call an individual parameter that the user enters P (probably just as a numeric value otherwise they'll be swamped with sliders), then the modified parameter is M_p. So M_p = ((P-1)/10)*W + 1.
Thus M_p is 1 when W is 0, and P when W is 10.
The full equation then becomes :
V = M_i * M_p * M_a * M_t * M_b * M_e * M_r
And the maximum possible V value becomes, letting W_x be the weight value for each parameter, W_i * W_p * W_a * W_t * W_b * W_e * W_r.
Of course there's no particular reason that 0-10 values have to be used, it just seems like a nice intuitive range. For sliders the default value could be shown as 0 for neutral with positive and negative values for more and less importance, but the actual numbers used could always be positive as above.
And thus political disasters are averted. :P
I might be misunderstanding something, but I have the hunch that you reinvented Multiple Criteria Decision Analysis (MCDM). ;-)
ReplyDeleteNot that the area couldn't use some solid software tools. The state of the art is atrocious.
Rhys Taylor Yep, that's the only sane approach. I'll be scribbling away, tonight. I'll gribble up something.
ReplyDeleteOops, I got the the maximum possible V totally wrong... if it was just the multiple of the weights, that would be zero if a single weight is zero. Whereas one of the nice things about this is that that doesn't happen : if you think an issue is totally unimportant, it won't reduce V to zero even if you completely disagree.
ReplyDeleteSince the maximum P can always be 10 but W_x varies, the maximum V will be the multiplication of all the M_p's, so :
Vmax = (0.9*W_i + 1) *( 0.9*W_p + 1) * (0.9*W_a + 1)... etc.
Joerg Fliege Yeah, could be... I'm just making stuff up as I go along, which seems to be working well* for major Western leaders lately... :P
ReplyDelete* It isn't, but they're doing it anyway.
There must be tonnes of more sophisticated approaches to this, and it's surely only a crude approximation of how real decisions are formed. I think the multiplicative nature probably works well for ability and trust, but it might be more linear for the others. But it'll do for a start. None of the major "who should I vote for ?" websites seem to take any of these other values into account at all; they are all about policies.
Probably some criteria also need more emphasis as either a decision making or analysis tool, otherwise people will end up assessing the criteria in different ways. For instance the overall talent present in the MPs counts for little if the leaders and cabinets are a bunch of twerps who can't utilise it correctly. Trust is also highly individual-dependent. So I should state somewhere that one should assess these criteria keeping in mind who one believes each party would actually put in power.
I had a quick bash at this in Google Sheets. Of course, I have no idea how to do fancy stuff like customising labels on bar charts, much less making anything properly interactive. ;)
ReplyDeletedocs.google.com - Political Drake Equation
But it seems to basically work. Since my personal values have one or two very low values for most parties, the weighting makes not so much of a difference unless I de-weight those attributes. But it's doing more or less what it should be doing, I think.