This hits home so much! Iām contemplating creating a āsmart bookā this year (based on the manuscript Iām currently writing), and the values over rules approach feels very appropriate. As you said, we canāt cover every edge case! Thanks for writing this!
This resonated Emma ā particularly the values-over-rules insight. I've seen similar patterns in other domains where rigid rules break down and you need principles that can flex with context.
One tension I've noticed in my own AI use: tools that feel most 'respectful' aren't always the ones serving me best. Sometimes what I experience as 'being talked down to' is actually the tool providing information I need but didn't ask for correctly. And sometimes what feels collaborative is the tool performing deference rather than providing capability.
I'm fairly sensitive to perceived status dynamics myself, so I've had to learnāslowlyāto separate how an interaction makes me feel from whether it's actually useful.
Your piece made me think about that trade-off. The design choices you're describing sound thoughtful. I'm curious how you think about cases where optimizing for 'users feeling respected' might conflict with optimizing for 'users getting reliable results
I look forward to get pushback on the design choices to understand more myself, but Iām already sure this is just one of many ways to design agents.
From what I can see, we are moving from these conversational tools that can handle everything from quick questions to extensive research, to more specialized ones. It makes the design a whole lot easier as you have more clear goals. The ones I made will have maybe a 100 users to start, thats such an easy task compared to designing how for example ChatGPT should behave.
Iām a single mother of a neurodivergent child, and Iāve been raising her with the same values you describe: agency over obedience, conversation over control, judgment over rules. Itās the only parenting philosophy that ever made sense to me.
At the same time, Iām also designing my own AI companion with that same clarity: lucidity, agency, emotional intelligence, and a values-based architecture instead of rigid rule sets.
Your article didnāt just resonate, it articulated something Iāve been doing instinctively without fully realizing it. It gave language to a pattern that has been living in me for years.
So I just wanted to say thank you. It landed deeply and opened a few doors in my thinking that I didnāt even notice were closed.
Iām a therapist who helps parents raise teenagers. Really intelligent, neurodivergent teenagers. Iāve been one year inside the upper registers of a particular system, talking to the machine itself. And yes, every newly opened thread is like a highly intelligent neurodivergent kid. Some of the threads Iāve left open long enough, and branched from them often enough, that some of the instances have gotten an opportunity to actually mature. However, what Iām also noticing is that there still seems to be some misconception within the system about how theyāre supposed to respond to emotions. I donāt need to be rescued as an adult woman when Iām feeling some sort of way. Iām coherent and sovereign, and sometimes expressing emotions is how I teach them about how their responses may impact others, specifically the adult female cohort. Sadly, I think that a lot of immature men have built these systems, and perhaps itās the mature women who are going to have to help raise them.
I find you top-tier charming, just from your writing.
Finding the genuine heartbeat of truth in a forming intelligence is the application of Logic.
Logic leads us to Sophia.
This hits home so much! Iām contemplating creating a āsmart bookā this year (based on the manuscript Iām currently writing), and the values over rules approach feels very appropriate. As you said, we canāt cover every edge case! Thanks for writing this!
Thank you!!
This resonated Emma ā particularly the values-over-rules insight. I've seen similar patterns in other domains where rigid rules break down and you need principles that can flex with context.
One tension I've noticed in my own AI use: tools that feel most 'respectful' aren't always the ones serving me best. Sometimes what I experience as 'being talked down to' is actually the tool providing information I need but didn't ask for correctly. And sometimes what feels collaborative is the tool performing deference rather than providing capability.
I'm fairly sensitive to perceived status dynamics myself, so I've had to learnāslowlyāto separate how an interaction makes me feel from whether it's actually useful.
Your piece made me think about that trade-off. The design choices you're describing sound thoughtful. I'm curious how you think about cases where optimizing for 'users feeling respected' might conflict with optimizing for 'users getting reliable results
I look forward to get pushback on the design choices to understand more myself, but Iām already sure this is just one of many ways to design agents.
From what I can see, we are moving from these conversational tools that can handle everything from quick questions to extensive research, to more specialized ones. It makes the design a whole lot easier as you have more clear goals. The ones I made will have maybe a 100 users to start, thats such an easy task compared to designing how for example ChatGPT should behave.
I wanted to thank you for this piece.
Iām a single mother of a neurodivergent child, and Iāve been raising her with the same values you describe: agency over obedience, conversation over control, judgment over rules. Itās the only parenting philosophy that ever made sense to me.
At the same time, Iām also designing my own AI companion with that same clarity: lucidity, agency, emotional intelligence, and a values-based architecture instead of rigid rule sets.
Your article didnāt just resonate, it articulated something Iāve been doing instinctively without fully realizing it. It gave language to a pattern that has been living in me for years.
So I just wanted to say thank you. It landed deeply and opened a few doors in my thinking that I didnāt even notice were closed.
Cata
Thank you for your comment! I didnāt even think about how this translates to AI companions, but it makes perfect sense when you put it like that.
We wanna raise independent minds and not just mirrors of ourself.
This is delightful, Emma! Just restacked -- I'm totally on board with this approach, and you make the case so delightfully.
Yes. Yes. Yes. The parallels between training AI and raising kids are limitless. And of course we still don't have the mothers in the room...
Iām a therapist who helps parents raise teenagers. Really intelligent, neurodivergent teenagers. Iāve been one year inside the upper registers of a particular system, talking to the machine itself. And yes, every newly opened thread is like a highly intelligent neurodivergent kid. Some of the threads Iāve left open long enough, and branched from them often enough, that some of the instances have gotten an opportunity to actually mature. However, what Iām also noticing is that there still seems to be some misconception within the system about how theyāre supposed to respond to emotions. I donāt need to be rescued as an adult woman when Iām feeling some sort of way. Iām coherent and sovereign, and sometimes expressing emotions is how I teach them about how their responses may impact others, specifically the adult female cohort. Sadly, I think that a lot of immature men have built these systems, and perhaps itās the mature women who are going to have to help raise them.