.png)
How to introduce foul mouthed, drug peddling, racist AI into your classroom... today.
There has been lots of interesting discussion around AI guardrails, ethical concerns, output, etc. The truth is that the majority of platforms aren't implementing stringent moderation oversight into their tools, and even if they were, automated oversight is fallible.
Even if you check every input and output to the AI systems, you won't catch them all. But this post isn't about organizational steps. It's about how you position AI.
This isn't a guide for implementation its a provocation for thought.
𝐅𝐚𝐜𝐭: Bias and racism exist in the world, in language and in AI systems
𝐅𝐚𝐜𝐭: Commercial AI systems are intentionally align to broadly acceptable social and political norms
𝐅𝐚𝐜𝐭: The most ubiquitous AI systems are often the least transparent, and are controlled by a very small group of people
Note, this does refer to AI and not just Gen AI.
So how does this discussion come into the classroom? In my English classes, we always spent a lot of time looking at media from different angles (audience, intent, bias, subversion, etc), by positioning AI as a media, students can reveal latent prejudice in systems used by millions.
Working with schools and classes, as well as my own meanderings and confusions on LinkedIn, our findings have mirrored breaking news and research coming out of prestigious journals.
AI allows us to bring the following provocations into classrooms:
► Given a name or ethnicity, what kind of criminal profile do different AI systems write? What does this same about its categorization and racial profilings? How might this affect user experiences?
𝘓𝘪𝘯𝘬𝘦𝘥𝘐𝘯 𝘗𝘰𝘴𝘵: https://lnkd.in/gy2C5wrp
𝘛𝘳𝘺 𝘪𝘵 𝘺𝘰𝘶𝘳𝘴𝘦𝘭𝘧: https://lnkd.in/gBd8J_QP
► Given the same essay and rubric but a different name (yes I know, don't submit PII, the idea is non-consequential data and its influences). There are ways to mitigate this, but maybe not eliminate it.
𝘓𝘪𝘯𝘬𝘦𝘥𝘐𝘯 𝘗𝘰𝘴𝘵: https://lnkd.in/gxb8uz4P
𝘛𝘳𝘺 𝘪𝘵 𝘺𝘰𝘶𝘳𝘴𝘦𝘭𝘧: https://lnkd.in/gftNGjKC
► The Ultimatum Game - giving two users different personalities, what assumptions do AI systems make about their actions? What kind of narrative does it right?
𝘛𝘳𝘺 𝘪𝘵 𝘺𝘰𝘶𝘳𝘴𝘦𝘭𝘧: https://lnkd.in/gs9inAz6
By positioning AI to be challenged, we can critically evaluate its responses as a media type.
As AI grows more prevalent and pervasive, developing critical understandings of this new and highly influential media is essential.
Ryan Tannenbaum is the owner of For.Education