What features would you like a prediction platform to have?

post by Mati_Roy (MathieuRoy) · 2020-10-13T00:48:03.024Z · LW · GW · No comments

This is a question post.

Contents

  Answers
    6 Daniel Kokotajlo
    5 ChristianKl
    5 Mati_Roy
    2 Mati_Roy
None
No comments
undefined

Answers

answer by Daniel Kokotajlo · 2020-10-13T09:22:43.531Z · LW(p) · GW(p)

--Ability to draw probability distributions with a mouse or pen rather than having to input numbers --Ability to (optionally) auto-update predictions over time, e.g. "I predict 30% chance of this happening by end of year 2021, if it hasn't happened yet, consider my prediction to be proportionally lower." A version of this for distributions would work as well. --Ability to compare a set of users by collecting the set of questions they all predicted and seeing who did best on that set --Massive scale: Thousands of questions on hundreds of topics, with some sort of recommendation algorithm to help match me to questions I'm interested in. Lots of users too, to make for lively discussions.

↑ comment by Mati_Roy (MathieuRoy) · 2020-10-13T23:12:55.195Z · LW(p) · GW(p)

if you predict updating your temporal distribution after 2021 conditional on it not having happened before, then you already have: the shape of the curve after 2021 shouldn't change just based on learning that it hasn't happened before 2021

answer by ChristianKl · 2020-10-13T11:42:29.291Z · LW(p) · GW(p)

There are four kinds of predictions:

  1. Public predictions
  2. Predictions under friends 
  3. Predictions within institutions
  4. Personal predictions

When creating a platform it's worth thinking about which use-case you want to focus on. Metaculus seems to me like it does a good job at public predictions. 

When it comes to prediction under friends I would both want the ability to share a prediction with specific people and in the past we found software that calculates the fair payoffs when either side wins the bet to be useful.

 For predictions within institutions, role and access management is important.

For personal predictions, UI to quickly enter data is essential. 

answer by Mati_Roy · 2020-10-13T00:58:04.976Z · LW(p) · GW(p)

updated: 2020-11-07

0) basic features

a) username, login

b) probably other basic things I take for granted

1) oriented towards long content

a) exportable, ideally in an open source format (ideally not just your predictions, but everyone's predictions)

b) archivable, notably on the Wayback Machine

c) organizational infrastructure and commitment to keep the prediction platform up for the very long term

d) notifications for when a prediction is due (if I'm the one that has to resolve it) or notification for when a prediction is resolved

2) able to input predictions in various formats

a) notably distribution over time

3) able to share a page summarizing my predictions, ideally with the option to have a comment on each

a) categorizing predictions

4) community features

a) moderators reviewing public questions

b) able to predict on all public questions

c) have a point system that allows never closing questions (unlike Metaculus)

d) able to comment on predictions, reply to comments, tag users, and adjust notifications

5) advanced predictions

a) able to use the output of some predictions as input for other predictions; maybe in a way similar to Guesstimate

b) able to use others (specific or aggregated) predictions as part of a function for your own prediction

↑ comment by Mati_Roy (MathieuRoy) · 2020-11-03T12:03:38.523Z · LW(p) · GW(p)

2) a) probability mass distribution over time and some other value

I would like to easily be able to predict on a question such as "What will be the price of neuropreservation at Alcor in 2030?" but for many years at the same time.

I was thinking what would be a good way to do that, and here's a thought.

Instead of plotting probability mass over price for a specific year, we could plot price over years for a specific probability.

So to take the same example, the question could become: "For what price is it 50% sure that Alcor will charge more than it over the coming century?" You could repeat the same question for the 10% and 90% mark.

Or you could just a specific distribution, like a normal distribution. And then you have just 2 curves to make:

  • What's the mean of Alcor's expected prices over the coming century?
  • What's the standard deviation of Alcor's expected prices over the coming century?

That way, you can easily get a probability mass distribution over price, over time.

x-post: https://www.metaculus.com/questions/935/discussion-topic-what-features-should-metaculus-add/#comment-44545

No comments

Comments sorted by top scores.