We launched Population summary in Elicit. This column shows more details about the population studied in the paper.
Compare this to the previous Population characteristics column - Population summary is so much more specific!
This is our first "compositional" column. Instead of answering 1 question about the population, Population summary aggregates the answers to subquestions such as Age, Organism, Country, and Sample Size.
With compositional questions, we can do exciting things like:
1. Integrate information from many parts of the paper
In the example below, Elicit pulls from 4 different sections (called Study Design, Characteristics of the Study Participants, Abstract, and Randomization and Study Groups).
Without a compositional approach, you can only get information from one of those sections. You would have a less complete view of the population studied.
2. Easily check Elicit's work
Despite using more information, a compositional approach preserves your ability to check Elicit's work. You can click each subquestion to see where Elicit gets the answer from. If we get a lot of thumbs down 👎 feedback on a particular subquestion, we can focus on improving those specifically.
3. Scale and customize
Lastly, a compositional structure makes it easy to customize answers for your specific needs. In the future, you can decide what subquestions are most relevant for you!
Decomposing complex reasoning into easier subquestions is core to Ought's mission. This talk from last Friday explains our research agenda and how we're breaking down even harder questions about placebos and adherence rates. If you want to learn more about our research and other organizational updates, you can sign up for the Ought newsletter here. (You are currently subscribed to the Elicit newsletter, which focuses on new features in Elicit.)