Using Gemini to Analyze a Conversation

I was listening to a Hard Fork episode, a new (to me) tech podcast from the New York Times with Casey Newton and Kevin Roose, where they had gotten responses from listeners on the good and bad ways they've been using AI in their jobs.

One of the responses talked about how they would put past conversations in to AI and ask for ways to improve or fix the interaction, which made me want to try that with an email exchange I'd had with someone. I had been trying to understand another person's concerns or fears over what was going on in a church community, and how this person's actions might have affected what happened. The exchange just kind of fizzled out and, in my view, I wasn't able to get any sort of satisfactory response to my questions.

So I copied and pasted the emails into Gemini, Google's AI bot previously named Bard, since I get access to that through the Google Workplace I pay for already.

The Prompt

I fed it this as a prompt:

Summarize this exchange between two people, [Name Redacted] and Chris, and tell me how it could be handled to diffuse conflict and have greater understanding:

And then pasted in the email exchange without email addresses included.

The Result:

Diffusing Conflict and Achieving Understanding:
Focus on Specifics: Chris seeks specifics about the concerns with the sermons. [Name Redacted] could answer directly about the content or avoid labels like "progressive Christianity."
Transparency: Chris desires clear explanations. [Name Redacted] could directly address the pastor's removal and its connection (or lack thereof) to the concerns.
Respect Preferred Communication Styles: Chris prefers written communication. [Name Redacted] could answer his questions via email while offering an in-person follow-up discussion.
Active Listening: Both could acknowledge each other's viewpoints and avoid assumptions.
By focusing on specifics, clear communication, and respecting preferred styles, they can move towards a more productive conversation.

The response seemed like Gemini was generally in support of my viewpoint, so obviously I'm happy with the answer. I felt like I had communicated my desire for transparency and openness in an email exchange, and that an in person meeting was certainly possible at some point, I wanted to have a greater understanding of what had happened before we met in person where emotions or reactionary conversations don't allow for a well thought out, considered response.

The "avoid assumptions" advice is interesting because I had referenced my desire to "...avoid making assumptions, but here's how I'm interpreting what you said..." in my side of the conversation.

The response from Gemini also feels a little like cliche advice, and are things that I generally strive for in communication, particularly when it's a written exchange vs an in person conversation.

But to someone who's doesn't feel safe to be transparent, being specific in details, or value the desire to take time to respond instead of reacting quickly, it could be an eye opening thought that would help inform future conversations.

I do appreciate the idea of submitting conversations like this to a neutral "authority" for analysis, even if it's an emotionless AI machine learning bot. As far as I'm aware, Gemini didn't know that the "Chris" in the email exchange was also the "Chris" submitting the prompt. But it's also not clear it didn't make that leap.

And I certainly wouldn't rely on AI to be the sole validation of my methods. But it is an easy way to augment feedback I might get from a close friend who could be much more partial to supporting my views or opinions, and less impartial.

Privacy Concerns

To be clear, Gemini doesn't (yet) read your inbox. I asked how I could submit an email exchange thinking that the emails are in my Gmail account and Gemini is owned by Google, they'd be able to just slurp it in, but Gemini said:

I currently cannot directly upload email messages for you to summarize. This is for security reasons as I don't have access to your personal emails.

I really hope that it stays that way, but I don't trust Google to maintain that wall between AI and my inbox. The amount of data they could train Gemini on if they had access to the world's inboxes has to be so tempting for Google.

For now, I'll use it sparingly on conversations and see if there's things that I can learn to help better communicate with people around me.

How are you using AI? Has your workplace had conversations around it's use, or is it up to each person to figure it out?