Skip to main content
Chat 1.0 is great for surfacing for quick and specific quotes, and looking for outliers. For more robust and comprehensive analysis we recommend using Chat 2.0. CoLoop’s AI chat works like ChatGPT but based on your research material, it’s effectively like having perfect photographic memory of all of your research material. You can use it to:
  • Find supporting evidence for a point
  • Stress test hypotheses
  • Summarise the main themes around an idea
  • Summarise entire interviews, activities or tasks

​
Creating an AI Chat

Click create chat in the LHS menu to create a new chat:

​
Asking questions in the chat

To ask a question in the chat simply type it into the message box and press enter:

​
Seeing supporting evidence

Click on a sentence in the generated answer to see the related evidence / direct quotes from the transcript.
  • This will open the evidence panel on the right
  • Darker sentences are more closely related to the sentence

​
Asking about specific participants segments

If you haven’t created a segment yet click here to find out how
You can also ask quesions about specific segments in your data. This is useful for things like comparing between different segments. To do this:
  • Press ”@” while typing a chat message OR press the ”@ Segment Button”
  • Select a segment from the list
Segments should be worked into your sentence so they make sense i.e.
✅ “What did @Dylan B. say about the electoral college system”
❌ “electoral college @Dylan B.

​
Filtering in Chat 1.0

After clicking create you’ll be prompted to choose a scope for the new chat thread. The scope can filter the chat to look at different subsets of your research material. This is particularly useful if you are:
  • “All” to search over everything
  • “IDIs / Focus Groups” to only focus on text transcripts, video and audio files
  • “Community Data” to only focus on xlsx or data uploaded from asynchronous qual tools
  • “Specific Files” to choose from a list of different files or even specific tasks and activities in your community data
image.png

​
When to filter the chat

Searching over everything is great however in some cases you may want to limit the scope to improve accuracy or target certain segments. Creating filtered chats is useful if you are:
  1. Conducting online qual and want to interrogate a single activity or task
  2. Working with focus groups and want to target a single one of them
  3. Conducting a mixed project with an online qual component followed by some interviews

​
Resetting the chat

  • The AI chat answers your questions using a mixture of the previous chat history and retrieved evidence from your research material.
  • In some cases this causes the chat to rely on things that were said previously in the conversation.
  • This can be confusing as in some cases it can cause drift, i.e. if the chat makes a mistake in a previous message, this will get taken into context in the next one.
  • Resetting can be helpful to minimise the impact of this