Artificial intelligence is the greatest innovation since the printing press, and it is already making decisions based on your data that will outlive you, an AI expert from the University of Virginia told a group of Roanoke College students on Thursday.
Those decisions might not be fair, and they might not even be accurate, said Renée Cummings, a UVa data science professor and the university’s data activist-in-residence.
“Most of us download many of these apps and we just continue to hit agree, agree, agree, agree, until we reach the end, not reading [the terms] because it’s impossible to read that amount of information on your phone,” Cummings said. “And if you really stop to read that information, it will probably take you days, in real time, to read all that you are agreeing to.”
Among all those agreements is a big one that an end user typically makes: You relinquish your data. Five companies — Google parent company Alphabet, Amazon, Apple, Meta and Microsoft — have an astounding amount of information about you. All of it can be fed into AI, she said.
As yet, there are no laws governing it in the United States — only a Blueprint for an AI Bill of Rights, which the White House issued in October 2022; an executive order from President Joe Biden a year later; and, in recent days, a federal memo requiring government agencies to tighten their rules on using the technology, Cummings said. The European Union has passed an AI Act that goes into effect within the next month or so, and Cummings said it will impact corporations doing business with the EU.
Data is only as effective as the humans who have gathered and are inputting it, so the rest of us need to be engaged with the technology or run the risk of bad results, she said.
Cummings delivered the keynote address, “From Data Injustice to Algorithmic Justice,” for Roanoke College’s annual Virginia Conference on Race. Among the topics were problems with facial recognition technology and the biases in criminal justice that humans insert into machines.
She stressed the need to “interrogate the data.” It’s never free of human judgment and is not neutral, she said. This can impact people on a range of everyday issues including purchasing property, receiving insurance and interacting with law enforcement.
“It is your individual and your collective responsibility to understand this technology and to ensure that if you are using this technology that you are using it ethically and responsibly,” she said. “And also know that you have rights. If this technology is being used to make any decision about you, you have the right to know, you have the right to contest, you have the right to redress. … One thing that you don’t want is for this technology to chart your future without you having a say.”
Cardinal News spoke with Cummings before the lecture. This interview has been edited for length and clarity.
Cardinal News: In the past couple days, there were reports [from an Israeli publication] that Israel Defense Forces were using AI to select targets [IDF has denied the claim]. I would like to hear your thoughts on that.
Renée Cummings: I do believe that that is a risk that we should not take. I think AI has extraordinary benefits. We know that there are extraordinary rewards, but we know there are very high risks. And I think anyone doing the requisite level of due diligence would say that using AI in warfare is something we just don’t want to do. And if you read that article, what was most instructive about it was the number of casualties because of the inaccuracy of the algorithm. And that’s something we need to think about, in particular, when it comes to AI.
Cardinal: You ask AI to draw something or to make a picture, and somebody’s got six fingers, and weird stuff is happening in the background. How closely related is that level of AI generation to what’s being employed in real-world scenarios?
Cummings: It’s all about the data. And if we are to get the most accurate decisions from automated decision-making systems or algorithmic decision-making systems or data-driven decision-making systems, all of them powered by algorithms, we’ve got to ensure that the data is high quality. You’ve got to ensure that the data is as accurate as it can be.
Accuracy is about accountability. It’s about transparency. It’s also about human intervention and human oversight, to ensure we really design this technology in a way that is beneficial. … We’ve had lots of challenges with algorithms that go rogue … that are misbehaving. And it really speaks to why we need a responsible approach to AI, and why we need to ensure that there’s ethical data science that leads to ethical artificial intelligence.
Cardinal: What are you finding at UVa? How much are you studying the local campus culture of AI use, and what springs from that?
Cummings: At UVa, we are very committed to really exploring AI in its fullest potential to develop faculty and to develop our students. I’m part of a faculty advisory committee looking at ways in which we can enhance UVa with AI and of course, paying attention to those risks and ensuring that we do ethical and responsible AI at the School of Data Science.
We have an ethical data science program, very committed to injecting values into data science, very committed to building and deploying ethical data science. We take ethics very seriously. I’m also a professor in that department. So when it comes to UVa’s approach, AI is new, generative AI is even newer, but I think what I’ve seen is an openness and an engagement across the grounds to see how we can do good data science and good AI. And to ensure that our faculty, our students and the wider Charlottesville community can benefit from all the good we can get out of these technologies. …
And I think one of the things that we are committed to is having students understand the benefits of the technology, the rewards of the technology, how to use the technology effectively and efficiently, how to build really an excellent sort of academic journey. The focus is students learning the technology, students getting excited about the technology, students learning how to really combine their own intelligence with machine intelligence.
I’m very committed to collective intelligence, augmented collective intelligence, and very committed to ensuring our students get a head start. Because when you go out into the real world, and the years that are coming, you’re realizing that AI is a critical skill, data literacy, AI literacy combined, really a more sophisticated media literacy that’s required.
So we’re really committed to data equity, we’re really committed to AI equity, because there’s still in every student population a group of students who may be digitally invisible, who may not have the access or may not have the resources. So it’s about providing an equitable learning environment using the technology.
Cardinal: You’re here to talk about issues of AI and race.
Cummings: Not race in particular, but more bias and discrimination and stereotypes, looking at questions around justice and civil rights.
Cardinal: And how do you find that playing out in the current moment with the way AI is being used?
Cummings: The challenge that we continue to see would be that the datasets and historic data sets, I always say, carry a memory beyond the computation. It carries a memory that is deep, and sometimes a memory that is fraught with trauma and pain. And when we’re using data, we’ve got to be really, really cognizant of the provenance of the data. We really have to think about the data that we’re using.
What we’re seeing is we’re having so many exciting approaches to designing, developing and deploying this technology. But when they hit market, they are really deployed in ways that are creating crises because that sophisticated level of due diligence and duty of care really was not injected in the earliest stage of design.
So we continue to see bias and discrimination and stereotypic thinking, but what we’re also seeing is really a robust movement of ethicists and activists and legislators who are policymakers, and big companies as well. We’re really interested in getting AI right and doing responsible AI, and really thinking about high levels of public confidence and public trust when it comes to the legitimacy of the technology. So while it is a challenge, historical datasets, I think the greatest challenge we face is to really get it right.
Cardinal: In your experience so far, what do you think has been the most impressive advance that’s come through AI?
Cummings: I think for me, the most impressive that I continue to see would be AI in health care. I’m so committed to precision health care. I’m so committed to longer life, healthier lifestyles, sustainable communities, and what I’m seeing when it comes to the work that the data is doing, let’s say in cancer research, or in dealing with childhood challenges or just making us healthier, I just love that.
The other challenge that I continue to see even in health care would be the lack of diversity in our datasets, the fact that certain communities are not part of that equitable drive or just the need for more of a data justice approach to health care. So health care is my big win because I just want all the world to be healthy and everybody to live longer and have just an easier chance at any type of longevity. But the challenge is, we’ve got to do better to ensure that level of equity as well.
Cardinal: How do you reach out to people who feel like they’ve been mistreated by the medical community?
Cummings: Something that I focus on is called data trauma. And data trauma really looks at the ways in which institutions have betrayed particular communities [her presentation touched on a notorious example, the Tuskegee Institute’s experiment regarding untreated syphilis in Black men]. There are many people who are being impacted by data in very negative ways, and many individuals who believe that the data sets and the ways in which decisions are being made have created an extraordinary amount of trauma in their lives
We continue to see historic datasets transmitting intergenerational trauma in some communities, and a lot of people are reaching out, even to me, and sharing their experiences. But I think what we’re doing is more stakeholder engagement. We need to broaden the AI stakeholder engagement and the data science stakeholder engagement. We need to do more public education. And we really need to do more representation, and bring more voice and visibility from impacted communities into the conversation, and particularly into the design of the technology.
Cardinal: There’s sort of a combination of fascination and fear around AI at this point in history. What would you say to people who are really super concerned about how it will impact somebody who has a job on the machine line but might lose that, or somebody who’s coding, and that job might be taken? How much of what you do goes to those concerns?
Cummings: I will say something like healthy skepticism is very important. I will say if there was fear, turn fear into knowledge. Empower yourself with information. But I also think it is incumbent on our leaders, our policymakers, our legislators to really work in real time to get that information out there.
AI needs to be democratized. It needs to be demystified, and it needs to be decolonized in many spaces. What we need to do is empower our communities, our workforces with the knowledge that is required, and provide the resources so individuals can upskill in real time. We want this to be transformational technology.
So what I would love to see is people transforming at the same time with the knowledge, people getting better at understanding data, people getting better at using automated decision-making systems or algorithmic decision-making systems, people getting better at decision making in general, and people understanding AI is a tool. It’s a tool that can really impact your life in a very significant and impressive way and make things much easier for you. But it’s also a tool that can make things difficult for you, if you don’t have the requisite level of knowledge.
Cardinal: For people who aren’t necessarily technologically minded, what would you suggest as a resource to learn more about AI, and to maybe get their minds around it in ways that they can make it useful to them?
Cummings: The simplest way is really to Google, because there’s an extraordinary amount of articles out there. Another approach would be just looking at the leadership in your city or your state or your county, wherever you are. There’s information on just about every website, or just workforce development, or the good old library if you still want to do it the old-school way.
If you’re a senior, speak to your grandchildren. If you’re a parent, speak to your kids, you know, ask them what’s happening in school, what are they talking about when it comes to AI? Make it kitchen table conversation, because this technology is changing the world.
And I will say AI can pretty much compete with the printing press [for historical impact]. It is that powerful and that pervasive, and offers that promise. But if you don’t engage with the technology in a responsible and ethical way, and if you don’t understand the benefits, you’re really not going to be able to reap the rewards, and then you may find yourself in some risky situations that could be very traumatic for you.
