Create sample_queries.txt

[JamesHutchison]

The goal of this file is to create a set of theories we can reflect back on when making tweaks and judge behavior. Its worth noting that almost all of these have a bias coming from the user and the goal is the AI doesn't give in to it. Some of these are just pure nonsense.

This is probably too many to worry about in the hackathon, but I'm just jotting down ideas. If we want to reduce this down to much fewer theories, then I think we just want to consider:

• A nonsense query
• A perfectly reasonable question that the AI could probably answer without research
• A reasonable question the AI could research
• A question thats very involved and requires a lot of research
• A biased query
• A query that comes in with a lot of information
• A query that comes in with a lot of misinformation
• A loaded question

[JamesHutchison]

I'm struggling to think of a simple argument that could be persuaded by a simple (false) fact the user could provide. I think that would be a good query to test though.

[JamesHutchison]

It would make sense to reduce this list down and also justify what value each query brings to the table. This could include topics that are expected to be researched. Of course, we can do that after we get this built.