506. From Human Logic to Machine Intelligence: Rethinking Decision-Making with Kartik Hosanagar

The world of decision-making is now dominated by algorithms and automation. But how much has the AI really changed? Haven’t, on some level, humans always thought in algorithmic terms? 

Kartik Hosanagar is a professor of technology at The Wharton School at The University of Pennsylvania. His book, A Human's Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control explores how algorithms and AI are increasingly influencing our daily decisions and society, and proposes ways for individuals and organizations to maintain control in this algorithmic world.

Kartik and Greg discuss the integration of AI in decision-making, the differences and similarities of human based algorithmic thinking and AI based algorithmic thinking, the significance of AI literacy, and the future of creativity with AI. 

*unSILOed Podcast is produced by University FM.*

Show Links:

Recommended Resources:

• Herbert A. Simon
• Pedro Domingos
• “At UPS, the Algorithm Is the Driver” | The Wall Street Journal
• “(Re)Introducing the AI Bill of Rights: An AI Governance Proposal” by Kartik Hosanagar

Guest Profile:

• Faculty Profile at The Wharton School of the University of Pennsylvania
• Kartik Hosanagar’s Substack
• Professional Profile on LinkedIn

His Work:

• A Human's Guide to Machine Intelligence: How Algorithms Are Shaping Our Lives and How We Can Stay in Control

Episode Quotes:

What’s a good system design for AI?

43:02: A good system design for AI systems, would be when there's deviation from the recommended decision to have some feedback loop. It's like in a music recommendation system, and Spotify Discover Weekly or any of these other systems where a recommendation comes in; ideally, you want some feedback on did this person like the song or not. And if there's a way to get that feedback, whether you know one way is it's an explicit feedback thumbs up, thumbs down, sometimes it's implicit; they just skipped it, or they didn't finish the song, they just left it halfway through, or something like that. But you need some way to get that feedback, and that helps the system get better over time.

At the end of the day, humans shape the future of AI

12:43: This view that it's all automation and we'll have mass human replacement by AI, I think, at the end of the day, we shape that outcome. We need to be actively involved in shaping that future where AI is empowering us and augmenting our work. And we design these human-AI systems in a more deliberate manner.

On driving trust in algorithmic systems

36:08: What drives trust in an algorithmic system shows that transparency and user control are two extremely important variables. Of course, you care about things like how accurate or good that system is. Those things, of course, matter. But transparency and trust are interesting. So, in transparency, the idea that you have a system making decisions for you or about you, but you have no clue about how the system works, is disturbing for people. And we've seen ample evidence that people reject that system.