Hi. It seems my original post about first-hand and second-hand data has already disappeared into the 'older than 90 days' void, but wanted to revisit this issue, since my team is asking about it. I initially voiced my concerns about second-hand data potentially biasing AI-generated summaries and limiting visibility into its origins. Now, I’d like to explore how both first-hand and second-hand data could co-exist in the platform. We have two recent examples where access to both types of data is valuable, though they should not be treated equally in AI summaries:
We receive feedback from internal support roles who help employees use our tools. This feedback is useful but considered second-hand, whereas direct input from daily tool users should be prioritized.
We conduct interviews with internal stakeholders about customer needs. These insights are helpful but not equivalent to direct customer data.
We expect more such cases in the future. Ideally, we’d like to store all this data in Dovetail and have it visible to everyone, since second-hand insights are often valuable for early exploration. However, the two data sources should be clearly distinguished in AI summaries and analysis. I’ve added a field to flag second-hand data in notes, but this doesn’t fully address the issue. Could this theme be considered in Dovetail’s development roadmap?
I actually use tags for this. Every single highlight is marked with the user type (we’re a two sided marketplace) OR “internal user” so I can filter for which type of data I’m looking for You could also probably do this with global fields on the data/note level. Would only work on business/enterprise though.
Pam D. For us, first-hand data comes from people who actually use the tool or service. Second-hand data comes from people who talk about those users (usually in support roles). These people often have great insights, but their input is filtered through their own experiences. In interviews, support roles can give both kinds of data:
First-hand: how they use their own tools or manage their teams
Second-hand: what they’ve heard or observed about others using the tool
That’s why Emi F.’s idea of tagging first- and second-hand data makes sense and I'll be adopting it as well 🙏 (if I understood it right). Most of our users just want quick AI insights and designers doing research don’t usually worry about the difference between data types, they just want useful info fast. That’s why I wish there were system-level tools to automatically tell the difference, so AI search could handle it better. Then I could still tag second-hand data when I spot it 🤓
so it sounds like you have the categorization of the data sources handled, but what is missing is the ability to prevent it from being incorporated into the AI summaries?
you'd hope that in the end, you could specify that in any of the chat features: "Separate out insights from first hand v second hand data sources. but it would also be nice to have that for the default summarizing features.
Right on the mark Stephen! 🎯 Would love if the AI summary could "evaluate" the data, so that it could say e.g. 'People in your sales organization are struggling with A, B and C, but their managers are also concerned about D and E being issues for them'. That would clearly give me an idea of what our users are actually saying, rather than just grouping topics A-E together and of equal value
