Certainly, here is a prediction market

Creating prediction markets is not easy. Here too, AI can help.

ChatGPT likes to use the phrase “certainly, here is” and there are plenty of scientific papers now, where it shows up. While that isn’t a good sign, there are plenty of good uses for large language models.

Forecasting is tough. AI is not that good so far. The money aspects of prediction markets make it even harder. You know what also is not easy? Creating prediction markets.

Puzzled robot illustration via Bing.

Creating prediction markets involves an understanding of the topic at hand and a knack for phrasing the market question in a way that is clear, concise, and unambiguous. It also requires you to clearly define the conditions for market resolution. This can be a daunting task, especially for those who are new to prediction markets.

Apparently, Manifold has not enough good markets yet, so they now have a partner program to incentivize creators.

I don’t qualify for Manifold’s partner program, because I’m not good at creating popular markets. Take that as a disclaimer for this post.

This wiki collects templates and also some generic advice. Some resources exist for those willing to learn.

Other people would rather pay someone to create markets for them. For example, here is a bounty for prediction markets about conlangs.

In between those two approaches there is another one. What about using that new thingy everybody talks about: AI has emerged as a powerful tool in various fields. It can help streamline the market creation process, making it more accessible to a wider audience.

It works fine.

Bing Chat works well because it uses a web search in the background (in contrast to the free-tier ChatGPT). Let me spell it out:

1. Describe your topic/question/idea.

2. After that paste the following:
   Write a prediction market about that using the following pattern. Write engaging but terse. The pattern:

   Title: <my question in YES/NO style>

   <a paragraph of context with more informative links>

   Resolves YES if <carefully considered conditions which are easy and objectively to decide>

   Resolves NO If <carefully considered conditions>

   <statement about the close date>

3. Let the AI write the prediction market text and copy-paste it to Manifold.

4. Some slight editing might be necessary. Details matter. Add a link or two.

AI has the potential for creating prediction markets. By semi-automating the market creation process, AI may help to democratize access to prediction markets, allowing more people to contribute and widen the topics.

---

INFER (now owned by RAND) published two things: Q&A with two Pro Forecasters and national security experts and How They Do It: Lessons Learned From Our Top Forecasters.

On Metaculus a post by Joker dives into the depth of forecasting birth rates and what drives them.

Exactly one year ago, the first post of the german Marktweise newsletter appeared. If you already follow me for a long time, how did it change over time? As a reader you probably look at it very differently than I do. Tell me anonymously.

Probably until next week, lazy readers! 😊