Welcome to the Community

Please choose one of the options below to log in and get stuck in!

Login with PPUK

Randomly sample manifestos to avoid bias

Dear Party,
    I'd like to share with you a manifesto sampling technique that I devised before the 2010 general election. As many of us will not be able to vote for our party in our local constituency, you may find this useful for choosing an alternative.
    It is a means to avoid some forms of psychological bias, in that we're more receptive to what we're already biased towards, and likewise ignorant to what we've not been exposed to. It is a scientific approach to processing the information, using random sampling to obtain a fair sample.
    If more people participated in this random sampling technique, it would effectively increase society's ‘immune system’ to circulating misinformation. People that have read a part of a manifesto in detail can effectively become a ‘misinformation antibody’ for that part. The more people that happen to know, in detail, about each part of each manifesto, the less a myth can propagate before it gets corrected by someone who has actually read-up on this matter. On average, the randomness will spread participants' collective attention evenly across the manifestos, and will uniformly strengthen this misinformation immune system. This is the best way that I can think of to cut through propaganda.
    The nice thing about it is that it's scalable to however much time you wish to spend reading manifestos, whilst still presenting you with a fair sample of the manifestos. You can stop at step 4 if you wish, having read only 1 topic in 1 manifesto. Or, you can read multiple topics, read multiple ‘rounds’ of topics, or even use it to read the entire lot of manifestos in a randomly structured order that will help reduce the effect of some forms of psychological irrationality resulting from the order in which you'd naturally read them given your existing bias.



---- The method ----

Step 1.
• Obtain the manifestos of the parties of the candidates for your constituency. Preferably get all of them for a fully unbiased sample.
• Get a die, or failing that, a coin.

Step 2.
• Randomly select a manifesto. The simplest way to do this with a high randomness is to roll the die or flip the coin once per manifesto, select the ones with the highest number, or heads, and repeat until you are left with just 1 manifesto.
• Note that you must commit to the randomness before the roll or flip – if you don't commit but instead pick and choose which dice rolls you like, then it loses randomness and becomes biased.
• This manifesto becomes your point of reference for this iteration (this topic).

An example of how this is done for 5 manifestos, A, B, C, D, and E:

With a 6-sided die:
A: 3
B: 1
C: 2
D: 6 ← D is the reference manifesto for this round.
E: 5

With a coin:
A: T
B: H, T, H, T, T, T
C: H, T, H, T, T, H ← C is the reference manifesto for this round.
D: T
E: H, T, T

Those were real examples that resulted from rolling a die that I have and flipping a twopence coin. As you can see, using a die is typically far quicker. Using just 1 die or coin means that any independent bias in the die or coin applies equally to all of the options.[note 1]

Step 3.
• Using the previously selected manifesto as the reference, open to the contents page and randomly select a topic using the same technique.

Step 4.
• Read some or all of that topic in this randomly-selected reference manifesto.

Step 5 (optional).
• Try to find the same topic in the other manifestos and compare to the reference. Don't expect to find the same topic everytime – the fact that there are different numbers of topics in the various contents pages alone means that there will be many topics that don't match up.
• If you do find corresponding topics then try to still use the randomly selected reference manifesto as your point of comparison for this topic iteration in order to avoid the bias that may arise from the psychological irrationality that the choice of reference for comparison can affect decisions. (This is the same psychological phenomenon that drives the decoy effect.)

Step 2a (as many times as you wish).
• You can repeat the previous steps 2–5 (or 2–4) any number of times that you feel you have time for. However, when repeating, exclude the previously selected references from the possibilities.
• If you've now done several topics, e.g. a topic from the contents page of each of the manifestos D, B, A, and E, then the next topic would be randomly-selected from the contents page of manifesto C as the reference manifesto. After this, you've completed a ‘round’ of topics and all manifestos are available as possibilities again for being reference in the next topic iteration.
• Don't feel the need to complete a round of topics; you can break-off after any number of randomly-selected topics, and, thanks to the randomness, you'll have considered a reasonably fair sample of the manifesto material. It's better that you consider at least 1 randomly-selected topic (1 topic iteration) than be put-off by wanting to complete a full round.
• If you've read all of the topics in a manifesto, obviously that manifesto can be excluded from being selected as a reference in step 2.



Note 1: Mathematicians reading this may be thinking of even quicker methods for randomly selecting from a larger set of possibilities (such as topics in the contents page) that involve base 6 or base 2 (i.e. binary) and prime factors to minimise the number of rolls needed, but these are more complicated to explain and rely more heavily on the fairness of the dice or coins. Provided that only 1 die or coin is used, the method that I describe only relies on the independence of die or coin outcomes, not the fairness.

Btw., a related point is that of search engine filter bubbles. Search engines like DuckDuckGo and Unbubble are especially important around elections.
    I hope that you find this information useful and I hope that telling people about this technique will improve society's immunity to misinformation. I hope that you can join me in trying to be a propaganda antibody.

Yours sincerely,
James Haigh.
Sign In or Register to comment.