Skip to main content
The Product Experience

Rerun: AI ethics advice from former White House technologist - Kasia Chmielinski (Co-Founder, The Data Nutrition Project)

31 min episode · 2 min read
·
Kasia Chmielinski

Episode

31 min

Read time

2 min

Topics

Health & Wellness, Investing, Startups

AI-Generated Summary

Key Takeaways

  • Product Management Trade-offs: Standard PM practices prioritize speed and ideal users, creating DNA-level exclusions. Early design decisions become permanent architecture, making marginalized users perpetually secondary. Building accessible-first or for edge cases produces better products for everyone long-term.
  • AI as Process Not Product: Treat AI development as multi-stage process including use case selection, training data, deployment, monitoring, and decommissioning. Build componentized systems to isolate and test each piece separately. Implement evaluations and red teaming at every stage rather than only at launch.
  • Vendor Procurement Questions: Before contracting AI vendors, demand answers on training data sources, accuracy measurement methods, ground truth comparisons, update frequency, and decommissioning criteria. Build accountability into contracts since customer complaints target you, not third parties, regardless of who built the system.
  • Data Nutrition Labels: Standardized dataset labels surface qualitative information like data cleaning methods, funding sources, intended uses, known issues, and ethical assessments. Organizations using labels report improved dataset quality because documentation requirements force better curation decisions upfront before model training begins.

What It Covers

Kasia Chmielinski, former White House technologist and UN adviser, explains how product management practices inherently create bias in AI systems and provides four concrete strategies for building more responsible technology that serves marginalized users.

Key Questions Answered

  • Product Management Trade-offs: Standard PM practices prioritize speed and ideal users, creating DNA-level exclusions. Early design decisions become permanent architecture, making marginalized users perpetually secondary. Building accessible-first or for edge cases produces better products for everyone long-term.
  • AI as Process Not Product: Treat AI development as multi-stage process including use case selection, training data, deployment, monitoring, and decommissioning. Build componentized systems to isolate and test each piece separately. Implement evaluations and red teaming at every stage rather than only at launch.
  • Vendor Procurement Questions: Before contracting AI vendors, demand answers on training data sources, accuracy measurement methods, ground truth comparisons, update frequency, and decommissioning criteria. Build accountability into contracts since customer complaints target you, not third parties, regardless of who built the system.
  • Data Nutrition Labels: Standardized dataset labels surface qualitative information like data cleaning methods, funding sources, intended uses, known issues, and ethical assessments. Organizations using labels report improved dataset quality because documentation requirements force better curation decisions upfront before model training begins.

Notable Moment

Chmielinski reveals building COVID vaccine equity systems that misclassified their own identity, demonstrating how technologists creating AI systems often fall into the gaps of their own products, becoming victims of the binary classifications and assumptions they programmed into algorithms.

Know someone who'd find this useful?

Episode Transcript

Hey. It's the Product Experience podcast, and I'm Randy. And I'm Lily. And it's been a while since we did an intro together. Yeah. We were lucky enough to join Pandemonium earlier this year, emceeing the Mind and Product stage while the other person interviewed people for the podcast. And it's been great to meet people face to face, but we've missed being together, so we promise to do more intros together. We met a lot of amazing people in North Carolina, but today's guest might have been my favorite. Don't tell any of the others. They're an adviser to the UN. They've worked with the White House. They were part of MIT Scratch, which might be my favorite part of the whole thing, and they also helped scale The US' COVID response. Kasia Kamalinski gave a great talk on stage then sat down with Randy to dig even deeper on data nutrition in this interview. Let's get right to it. Hey. We're here in Raleigh at Pandemonium, and I'm here with Kasia, and they just got off stage and gave this amazing talk. And we were hanging out yesterday and had some really good things to talk about as well, and we're gonna get really into all this stuff. But first, for anyone who isn't here and didn't get a chance to see you, can we just do a quick introduction? Tell us what are you doing these days and how did you get into this world in the first place? Yeah. Thanks for having me. It's great. The energy here is fantastic. Or for those who can't see us, we're surrounded by small bright pink dinosaurs, which I love. It's a good vibe. Yeah. So my name is Kasia. I am a technologist. I've been building products for twenty years, which is a scary thing to say. As a product manager, and I've worked in a number of different places. At this point, I'm really focused on responsible AI and data and ethics, and I do this as a consultant. So I work with a number of organizations, including the United Nations, and I also run a small nonprofit called the Data Nutrition Project. We build nutrition labels for datasets. Which is fantastic, and we will definitely get into all of that. Let's start with the UN though. Tell us a little bit about the work you do with them. So the UN is massive. So shorthand, I say I work at I work with the UN, but there's actually so many different components, and I'm still learning the system. I've been an adviser with them for about three years, I wanna say. Two different parts of the UN, both of which are really thinking hard about data quality, data standards, and then how to use data, either internal UN data or external third party data to build and deploy algorithmic systems that will benefit the sector. So what that really means is I'm I'm a team of …

Get the full transcript (6,565 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all The Product Experience transcripts →

You just read a 3-minute summary of a 28-minute episode.

Get The Product Experience summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

Keep Reading

Books, tools, and gear mentioned in this episode

SignalCast may earn commission on purchases via these links.

other

  • Data Nutrition Labels: Standardized dataset labels surface qualitative information like data cleaning methods, funding sources, intended uses, known issues, and ethical assessments. Organizations using labels report improved dataset quality because documentation requirements force better curation decisions upfront before model training begins.

More from The Product Experience

We summarize every new episode. Want them in your inbox?

Similar Episodes

Related episodes from other podcasts

Explore Related Topics

This podcast is featured in Best Product Management Podcasts (2026) — ranked and reviewed with AI summaries.

Read this week's Health & Longevity Podcast Insights — cross-podcast analysis updated weekly.

You're clearly into The Product Experience.

Every Monday, we deliver AI summaries of the latest episodes from The Product Experience and 192+ other podcasts. Free for one show.

Start My Monday Digest

No credit card · Unsubscribe anytime