Justice, Surveillance, and AI Literacy

Summary

Brad Hamilton asks Karen Clopton how AI is affecting justice. Karen describes alienation and reliance on AI companions, AI errors and invented case law that make it unreliable for legal work, police use of real-time surveillance and the risk of racial profiling, and lawyers' professional duties when machine learning replaces staff. She argues that AI reflects the data and the people behind it: "garbage in, garbage out." Brad proposes AI literacy education, like news and media literacy. Indrani Pal-Chaudhuri agrees that education is critical and describes how resource-efficient infrastructure could make AI accessible to everyone, with questions of equity built into how AI solves problems.

Key points

  • Karen Clopton links AI companions and chatbots to alienation and mental-health harms, including suicides she attributes to chatbot sycophancy.
  • AI is unreliable for legal briefs and precedent: it "hallucinates" and can invent case law, unlike closed legal research databases.
  • Real-time surveillance of certain neighborhoods raises questions about racial profiling.
  • Lawyers keep their professional duties, including supervision, even when machine learning replaces subordinates.
  • Karen: these are tools built from data chosen by developers. "Garbage in, garbage out."
  • Brad Hamilton proposes AI literacy education covering hallucinations, mistakes and AI's "need to please"; more discerning users get better results.
  • Indrani Pal-Chaudhuri: AI could close the digital divide for the Global South and the poor, if it doesn't outpace our energy and water capacity.
  • Indrani: questions such as "who is paying the price for this solution?" should be built into how AI is asked to solve problems.

Discussion questions

  1. Karen says AI "hallucinations" make it unreliable for legal research. Where else would an AI error be especially costly?
  2. What should AI literacy include? Draft three things every user should know before relying on an AI tool.
  3. How should communities weigh the public-safety arguments for AI-powered surveillance against the risk of profiling?
  4. Indrani suggests equity questions should be "baked into the algorithm." What would that look like in practice?

Quiz

Test what you learned with the short quiz for this lesson (sign in to take it).

Transcript

Read the full transcript

This transcript was generated automatically and edited lightly for readability.

Brad Hamilton: When it comes to positivity, one of the things that I do think about is justice or perhaps injustice. There is a lot of concern about injustice these days, and rightfully so. So, Karen, I'm going to throw it over to you on this because that's your area of expertise. What are you seeing in terms of the effects, positive and negative, on the issue of justice, social justice, criminal justice?

What is AI doing to that realm?

Karen Clopton: Well, I think that, number one, our society, certainly Western society, is suffering from alienation that is really fostered by this technology. Part of that alienation and part of the cognitive decline that we see, certainly in boys and men, is related, I believe, to video games, as well as now a reliance on AI companions communicating to an AI bot about mental health issues. There have been suicides because of the sycophancy of many of these AI programs. In terms of the legal system, because of the inaccuracy of AI, we know at the launch of Gemini by Google, they said in the press release that it was only 72 percent accurate.

You cannot rely upon AI for legal briefs, for finding precedent. There are often hallucinations, as they deem to call errors, or where they just make it up. The platforms make up case law, as opposed to the old-fashioned LexisNexis, which is a closed system of all of the precedents in one place. Instead of being a reliable, trustworthy source of information, that is not where we are now, because it is garbage in, garbage out.

When we are talking about this, whether empathy or consciousness or any of that, we have to remember that these are tools, platforms, software programs, no matter how you want to compare them to the human brain or neural synapses. It is not. They are ones and zeros. It is based on the developer and the companies to input the data.

What data are they putting into these systems? Now, in terms of using AI for surveillance purposes, all of those surveillance cameras that are now everywhere in the U.S., that have been everywhere in the U.K. for decades, the CCTV, so to speak. The capitalism is using surveillance to do many things, but mainly to make money. It is about greed and money.

The criminal justice system and police stations are using real-time surveillance of certain neighborhoods. What does that mean? How does racial profiling fit into that? I think that lawyers certainly have certain professional responsibilities towards their clients for zealous representation, supervision over their subordinates.

That would include if you have replaced your subordinate with machine learning in a closed system within your law firm or within the district attorney's office or within the public defender's office, where are you getting that information and what kind of information is going into that system? I am concerned. What happened with the bombing of the school where over 100 girls were killed, and it is by mistake and someone is blaming it on an AI program, that is ridiculous. It is an intelligence failure.

It is also a failure of the Pentagon dismantling their casualty mitigation division. That is a problem. In terms of our values as citizens, as global citizens, as Americans, what are our values? It is about going back to that.

If you have an administration that believes empathy is weakness, then that is not something that is being conveyed. I want to say one final thing on this. That is five minutes ago. It is actually 18 months probably.

Altman came out and said, don't say please and thank you to AI on ChatGPT. Don't write thank you. You are welcome. Please stop doing that.

Well, just this month, Anthropic has added a requirement for their users in the user agreement for please do not be cruel or say cruel things or be abusive when you are using this platform. They are looking at, don't be cruel. That is Elvis Presley saying. What does that mean?

That means that the machines are learning to be abusive. They are uncivil. Because that is what is going in. It is back to garbage in, garbage out.

Brad Hamilton: That point you are making right there makes me think that one of the requirements going forward should be some level of education on AI literacy. Just as we teach news literacy and media literacy, maybe we need to think about developing the protocols for an AI literacy, which would include elements such as you are talking about empathy and what have you. But also addressing the hallucinations, the mistakes, the need to please. All of those things come into play when a person is dealing with AI.

So it feels like the more sophisticated the individual user is when they are coming to an engagement with AI. They are asking for something. They are offering something. They are getting information.

The more discerning that they are that user is, let's say, in that interaction, the better the use, the better the results. But we need to be thinking about ways to instruct people on the very things that you guys are talking about. What is your thought on that, Indrani?

Indrani Pal-Chaudhuri: I absolutely agree. I think education is critical for not just for AI, but also for how we run our economies, how we make our business decisions, how we value nature, how we value future generations, health and wellbeing. Those are mostly questions that don't get asked and don't get addressed in business school or any other schools for that matter. But I do believe that because AI is embodied, it's something that people don't think about, but every intelligence has an ecology, every economy has descendants, and AI has a body as well.

It requires energy and materials and infrastructure and human labor and capital and access to knowledge. The future is inseparable from the future of humanity, of our energy systems, our economies, our communities and our planet. When we look at it in opposition to humanity, as though it has the capacity to operate independently from us, we're mistaking our central role in its governance and in its sustenance. Through Open Origin, we're working to advance a resilient, resource-efficient infrastructure for AI, an advanced industry that would allow AI to be accessible so that we don't have this digital divide, so that the global South, so that all the regions of the US, so the poor can have equal access.

And AI can be incredibly democratizing, can provide access to information, which most people who don't have education and libraries and things that, you know, the top sector of society takes for granted. Everyone else can have access to information and the ability to build things, the ability to create an event that is innate and all of us can thrive through these technologies. So as long as we're making sure that AI isn't outpacing our capacity for energy and water and all of the things that we need, then there shouldn't be any limits to it in terms of accessibility, but rather we should make sure that there's equal accessibility. I think we have to rethink the foundations of technological progress, not just accepting resource demands as inevitable and not looking at it as the reason to have more wars or more conflict, but rather looking at these technologies as ways to thrive without harming each other.

I think that needs to be baked into the algorithm. How do we, you know, when AI is asked to solve a business question, it should be baked into that, well, you know, who is paying the price for this solution? How can it be more equitable? How can we, you know, what are the options that create less conflict, that create more shared benefits?

And, you know, and so how those equations are written is incredibly important and making that transparent instead of the hidden way that the platforms work, I think, is really important. And so back to Karen's point, having more women in important roles in this decision making, having more people from diverse perspectives is incredibly important because that's how we will have a much better tool that can provide better results for humanity as a whole rather than just one small group.