Formulating Questions

A Hypothesis question is formulated according to a simple set of guidelines, the most important one being that it is asked in such a way that its resolution is binary. Other prediction-based platforms generally follow a modified version of this rule, mostly because predictions serve a different purpose for them than they do for Hypothesis. A good example of a traditional formulation is for a prediction about the potential 2020 Democratic nominee for President of the United States. A prediction market like Predictit would frame this question in a way such that it can accommodate many possible resolutions, that, while still offering “Yes” and “No” options for each possible solution, the question itself is framed in such a way that it serves as an umbrella for a variety of possible outcomes.

Hypothesis’ model would require for this question to be broken down into many individual questions, and in doing so change the way that the question is written. “Who will win the 2020 Democratic presidential nomination?” becomes “Will Joe Biden win the 2020 Democratic presidential nomination?” with another, separate prediction asking, “Will Andrew Yang win the 2020 Democratic presidential nomination?” and another asking, “Will Elizabeth Warren win the 2020 Democratic presidential nomination?” and so on.

In this format, a single question can only be resolved in one of two ways: either the event in question happens or it does not. Therefore, only two options exist for users to consider: “Yes” or “No.”

An important implication of this style of formulation is that it is incredibly simple, minimally intimidating to users and by extension more easily gamifiable. Your prediction of “Yes” or “No” will always lead to a resolution of “Yes” or “No” and you will always be either correct or incorrect.

Entrusting question formulation to the Hypothesis community is a double-edged sword: on one hand it releases Hypothesis from the editorial responsibility of producing unbiased, relevant content, but on the other hand it requires a strong structure for community-powered accountability that ensures the fulfillment of those same expectations.

First, we will return to the question about the 2020 Democratic nominee for a more detailed breakdown of the elements that go into formulating a question. The information contained within and immediately relevant to the question “Will Joe Biden win the 2020 Democratic presidential nomination?” can be described as follows:

Question: Defined as an actionable question with a future outcome. It always starts with the word “Will” and should be structured in a way that only allows for a binary outcome. We can examine two variations of the question “Will Joe Biden win the 2020 Democratic presidential nomination?” to understand how they are both valid within the context of Foresight’s system, but lead to two very different considerations by users:

Variation 1: “Will x happen?” maintains an open-ended resolution timeframe. Joe Biden could “win” the presidential nomination tomorrow or at the convention, but the fundamental question is whether Joe Biden will win the 2020 nomination.

Variation 2: “Will x happen within y timeframe?” sets an artificial resolution date, but we must be wary of the fact that this changes a fundamental part of the question being asked. For example, in rewriting the Joe Biden question to ask, “Will Joe Biden win the 2020 Democratic presidential nomination by July 1?” We no longer ask only whether Joe Biden will be the 2020 Democratic nominee, but rather if he will reach that status by July 1. A predicting user may believe that Joe Biden will be the eventual 2020 Democratic nominee, but does not believe that he will be anointed as such until after July 1. Accordingly, this user would predict “Yes” for variation 1, and “No” for variation 2.

Choices: Defined as the available user input to indicate a chosen prediction for a given question. These choices can only be “Yes” or “No,” as dictated by the rules of question formulation. Once a user makes a choice, they cannot change it.

Topic: Defined as the prevailing subject matter describing a given question. A given question can belong to more than one topic. Topics serve as a means by which to categorize questions so they may be filtered by users, as well as to provide users with more granular data about their performance over time.

Resolution: Defined as the outcome of a given question, which will always match one of the initial available choices of “Yes” or “No” .

Context: Defined as the content both generated in-app and submitted from out-of-app by users and groups that forms a knowledge base around a given question. There are no specifications limiting the acceptable formats of submittable research.