
If you’ve read my earlier blogs about Artificial Intelligence and teaching, you know that I’m convinced it’s here to stay, that the questions we should be answering are how can we use it to improve teaching and how can we limit the harm it can do. I don’t pretend to have the answers, but I’m committed to the questions and exploring our options.
Recently I had the privilege of reading Mason Broxham’s blog on Riddance called “The Accountability Gap.” As someone who believed that guardrails are our best hope to limit harm, I was startled to read, “At Riddance, we’ve observed a pattern in our investigations: When two or
more AI systems overlap, the guardrails designed to reduce harm and prevent Terms of Service (ToS) violations stop working like they should.” (www.riddance.ai) He goes on to explain that this is a structural phenomenon. Because AI is sometimes “stitching together AI scenes that were not necessarily problematic in isolation,” and because platform moderation is automated, AI is not capable of recognizing the harmful content that evolves from these patchwork efforts.
“The AI that made the content had no mechanisms in place to understand the harms it perpetuated, and the AI moderating the content had no general concept of what those harms look like in the first place,” he argues. Human beings make decisions in a context that informs those choices. Humans live in a “real, messy world of interconnected reasons and norms” that informs their “decision-making calculus.” AI models lack this kind of overall context, creating what Broxham calls the “accountability gap.”
This thunderbolt struck terror in my generally optimistic vision. If guardrails cannot create real accountability, what can educators do? I posed the guardrail issue to my husband and sons, all of whom work/worked in tech. In the AI arms race, one son sees a dire future while the other works on building better guardrails. My spouse argues that “humans using AI will probably be able to work around the above guardrails. Furthermore, immoral builders of an AI system can just not put in needed guardrails. All this means is that just like before AI, humans will be able to be immoral and illegal even using AI.” My pessimist son lists tools with capacities that keep me awake at night.
I struggle to maintain some optimism in the face of these discussions, and the inertia of education and the overwhelming power of AI make me wonder if we’re doomed. I don’t have an answer. I do know that we’re failing to ask enough questions, to bring together enough human minds to counter this trajectory. But we cannot afford to wait. I hope that people smarter about this than I am are doing more than I currently know… We need them to.



