Covert Contact 73: Data, AI, and Ethics

In this September 2017 conversation, entrepreneur Evanna Hu joins me to discuss our rapidly evolving relationship with data and its implications for national security. Evanna describes how her company, Omelas, was using data analytics and machine learning to understand communities and develop better ways to measure efforts to counter violent extremism.

We look at what social media, images, and natural language processing can reveal, and where that analysis runs into trouble. The most vocal people online may not represent a wider community, and recognizing extremist imagery is different from understanding the intent of the person sharing it. Evanna emphasizes the need to turn growing collections of data into useful analysis.

The ethical questions are inseparable from the technical ones. Evanna explains why Omelas was holding off on some commercially available datasets while it worked out its own standards, even when their use might be legal. We also discuss how assumptions and societal biases can find their way into algorithms, and why those limits need to be clear to the people using their results.

We close by considering the potential for these tools to support peacebuilding, along with the risks of manipulation and abuse. Underlying the discussion is a concern that people may not appreciate how much data they generate or how it can be used. Evanna argues for shared industry standards governing the collection and analysis of public information and the biases built into those systems.

This episode was originally released on September 5, 2017. Roles, technologies, and assessments reflect the discussion at that time.