From Wikipedia:

An influence diagram (also called a relevance diagram, decision diagram or a decision network) is a compact graphical and mathematical representation of a decision situation. It is a generalization of a Bayesian network, in which not only probabilistic inference problems but also decision making problems (following the maximum expected utility criterion) can be modeled and solved.

I'm proposing to use influence diagrams in an online platform which will mediate disagreements between its visitors. Here is how it may work:

  1. A controversial political decision is considered (e.g. How Brexit should be resolved?).
  2. Analysts of the platform create an influence diagram of the decision situation without specifying the parts which are controversial (e.g. How important is preservation of British identity?).
  3. Arguers of the platform list argument about how the controversial parts should be evaluated.
  4. Ordinary platform users explore the diagram, read the arguments and specify their opinions about the controversial parts similar to a questionnaire.
  5. Based on the user's inputs, the influence diagram recommends the decision with the highest expected value. Different users get different recommendations.
  6. Critical parts of the diagram which causes the most amount of disagreement are identified.
  7. Analysts review and provide more detailed models for the critical parts. Arguers focus on the critical parts to have the most influence on the decision recommendations.
  8. Steps 4-7 repeat until the diagram is so detailed and arguments are so comprehensive that the overwhelming majority of the participants have the same view of the decision situation.
  9. Either the decision which satisfies overwhelming majority emerges or the shared understanding is used to run a successful negotiation.

Can such a use of influence diagrams be effective? Is there a better way to use them in the context of politics?

Related questions:

1 Answer 1


This look to me like a fancy implementation of a plain debate, where people state facts, explain mechanisms and consequences and analyze them internally in their own view. There are advantages to political debates as with this process, as well as disavantages.

On the plus side

  • People may reach consensus on some things
  • People get a chance to change their view when they get the bigger picture

On the minus side

  • It can't be held at too many
  • People can be unequally represented e.g. analysts get power in skewing things
  • People would favor designing systems that advantages their view (as when people take arguments that favors their point of view)

But the most important reason not to frame debates into models, is that there are viewpoints that would disagree on the very nature of the models you would create. For example:

the influence diagram recommends the decision with the highest expected value

I would argue one should not model utility as a single value to maximize.

To sum up, I believe models can prove to be good arguments to explain a particular viewpoint and how it is built, but they can't compete with human decision when it comes to deciding for taking a side in a political disagreement.

  • When you say that models can't compete with human decision on a political issue, do you mean that they can't compete in popularity or soundness of their concussions?
    – Dante_ga
    Commented Oct 25, 2019 at 22:10
  • @Dante_ga Both. I don't think it's possible to make an accurate model of poll predictions (outside basing it on previous data), and I don't think it's possible to automate our executives. The thing is political decisions are human negotiations at their core, they call for human values that are, by definition, uncountable.
    – Diane M
    Commented Oct 25, 2019 at 22:25
  • Which definitions states that human values are uncountable? Please note that I'm proposing to adopt a decision recommendation tool which is already used in business. I'm not proposing to let it decide or negotiate something instead of its user.
    – Dante_ga
    Commented Oct 25, 2019 at 22:32
  • @Dante_ga These values en.wikipedia.org/wiki/Value_(ethics) are uncountable. I did note that you only want to recommend, but my point is that they are not recommendation that work well enough for being recommendation you can follow. I suppose they can work if you agree the model and data to be correct. This is what I explain in my summary.
    – Diane M
    Commented Oct 25, 2019 at 22:48
  • I didn't find anything in that article stating that those values are uncountable. On the contrary it includes a diagram which assigns numeric coordinates to different countries representing differences in strengths of different values.
    – Dante_ga
    Commented Oct 25, 2019 at 22:59

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .