Episode Transcript
[00:00:03] Margaux: Welcome to UVA Data Points. I'm your host, Margaux Jacks. In this episode, we explore the future of artificial intelligence through the lens of policy, ethics and governance, examining how this rapidly evolving technology is reshaping society and the responsibilities that come with it.
Joining the conversation are Renée Cummings, professor of Practice in Data science and a leading voice in AI ethics, and Mona Sloan, Assistant professor of Data Science and Media Studies, whose work focuses on the intersection of technology and society. Together they share insights on how we can guide the development of AI in ways that are responsible, equitable and grounded in public interest.
[00:00:50] Renée: Welcome to everyone who has joined us. And for me, there's no better person to have a conversation about AI with than my colleague Mona, who I love a lot, have an extraordinary amount of respect for, and she's certainly a leader in this space. So I'm going to ask you the million-dollar question, when do you use AI and when do you not use AI?
[00:01:12] Mona: All right, the million-dollar question, of course the answer to that is evolving as the technology is evolving. But I will give you my absolute favorite use case of generative AI, which I think is sort of the coolest and latest kit on the AI, your ChatGPT, your Claude, what have you. So generative a system that doesn't just predict future scenarios and make suggestions, but that actually generates new data and that
[00:01:42] Renée: is
[00:01:44] Mona: using those kinds of systems to generate Excel spreadsheets, which I very much need to run my lab and my life and operate. As a modern day scholar, I never learned how to actually write Excel formulas. Even if you do know that it takes time, what you can do with generative AI, you can tell the system what you need and you can task it with creating a spreadsheet and it does a really, really good job. So now I have a spreadsheet where I can keep track of my grants and I get sort of alerted when I only have a six month Runway or a four months Runway, which is something that would have taken me days prior, which is something that we didn't have available infrastructurally here in the department.
And I was just really grateful and happy when I managed to get my hands on that with the use of a generative AI system. Renée so that was a big, big AI moment for me. But let me actually throw it back to you.
We often joke that we use AI very differently.
One of us uses it for comparative analysis, the other one uses it to turn bullet points into prose.
Do you remember from our last conversation who is who?
[00:03:10] Renée: I think I do. For me it's always been about comparative analysis. I write things and I of course, ask the GPT to tell me, you know, what is it? You know, how does it sound? Is it good enough? What's missing? Rate it. So I do a lot of that and then I like to see what's happening globally and I like to compare articles and compare academic articles and journal articles and. And I do it for the, of course, for the comparative analysis, but also I do it sometimes for the bullet points because I like to see things in bullet points. I know you turn the bullet points into prose, but I think the technology, no matter how you use it, once you use it responsibly, offers us the kind of experience that we're looking for.
For me, you said Excel sheets. For me, I like business plans and plans, life plans and goal plans. I think it, it really does nice with the planning. The thing that I'm always concerned about is I'm just not going to use this along the knowledge route. So. So I'm really very circumspect and very critical as to what that information is. And nothing that it gives me, I take at face value. I read, I double read and I triple read. I always like different sources as well. So, yes, comparative analysis works well for me, rating my work, you know, being that sort of, what do you call
[00:04:34] Mona: it, that editor or devil's advocate, or
[00:04:37] Renée: devil's advocate, which is very important.
And it's really providing it provides the kind, I think, for whatever you need or whatever job you do, and whatever you're trying to engage it to do, it can pretty much deliver. But always with that measure of critical thinking and that understanding that what you're getting here may not be always what you really need at that moment.
[00:05:02] Mona: Yeah. And thank you for actually bringing that to the table so early in the conversation because I think what you're putting your finger on here is the role of AI as an epistemic technology. Right. It's a system that produces and co. Produces truth or claims to do so. And we know from the technological design that this is not quite accurate. Right. We're still sort of in the world of probabilities. We're not in the, in the realm of causal relationships. And so the double checking then really is profoundly important because the rounds of responsibility for truth and the production of knowledge is still a human responsibility.
Right. And so let me just throw another big question at you, Renée. So we sort of today sort of equate ChatGPT, you know, with any and all AI.
So from your perspective, is ChatGPT the end all be all of AI. Like, what is missing from that narrative, from that understanding, from your point of view?
[00:06:12] Renée: I will definitely say it's a very visible interface of the technology and it's a very strong brand of the technology.
I always tell people that we access many brands. We may not only have one brand of shoe or one brand of sneakers. So ChatGPT is a brand of this technology.
What really sort of gets me inspired or keeps me inspired is that we need to always go beyond the interface. I always believe that the tools are sort of like the tricks and the nice things that we can use and the entertaining things that we could use and the things that excite us. I always say this technology makes everyone feel more exciting, more creative, more intelligent, more and more. But it's also sort of masking the major challenges as we both understand this question of power and how powerful this technology is. For me, being a criminologist and coming from the criminal justice system, questions around decision making with this technology called AI, and the fact that the technology is being used to make some really high stakes decisions that we have not all decided that we want to participate in, or the question of invisibility, we're engaging with something that we really don't understand, many of us what's happening behind the interface and the question around irreversibility of many of these decisions that are being made. So yes, it is what we see and many people are engaging with the technology from the standpoint of that platform, but it's also creating a very exciting form of evasion as to what's really happening behind the technology. I always say it comes back to data. And There is no AI and no ChatGPT without data. The entire model is built on data. So the kinds of challenges that we continue to see with bias and discrimination or misogynistic types of approaches that we may be getting from the technology are things that we cannot forget. So while we explore the potential and we excite ourselves about the promise and we think about how amazing the technology may be making us feel, we've got to understand those challenges, those, those risks and the need for the protection and the governance and compliance and enforcement and all of those things, because this is just not a toy in our hands or a game that we're playing. This is the technology that is having extraordinary impact on the future of society. And as I bring in the future of society, I bring in that critical word, a social phenomenon. And we both speak about the impact of this technology, its social impact, and that's something you are extremely passionate about as well. Mona let's speak about that.
[00:09:05] Mona: Yeah, thank you for that. Q. Renée so as a social scientist, I look sort of beyond the sort of technical theater, if you will, and sort of I'm really more interested in the big social ideas and social processes that are embedded in any technology.
And what we can really observe now in this sort of world of AI that we all live in, is that AI has become what I call a social infrastructure. What do I mean by that? I mean by that that AI systems, which are predictive systems, they are designed to create predictions that then either humans or other AI systems act upon or that I've used to create new data, such as in the context of generative AI, that logic of prediction has become infrastructural and is baked into the digital infrastructures that we are already using to organize society, to organize work, to organize family life or intimate life, to sort
[00:10:09] Renée: of
[00:10:12] Mona: manage access to resources and institutions. For example, in this very moment, there is an AI system at work that is automatically transcribing what we are seeing in the background. The transcriptions can be switched on. There's also an AI system at work in the background that is actually making sure that the video is at least somewhat stable for everybody. So these systems are everywhere. We don't always see them.
And that is the case for all kinds of infrastructures. Infrastructures are in fact, designed to be in the background. We're not supposed to see the pipes, we're not supposed to see where the electricity comes from. We only sort of notice these infrastructures when they break down. And so the same thing is happening with AI. AI is everywhere. It is playing a sort of predictive role in how we relate to one another or how we, for example, get into college, get a loan, or gain any access to any other resource. But we only sort of notice this decidedly when it breaks down. And what this breakdown looks like is different for different kinds of communities. Going back to your earlier point, Renée, about the question, have we figured out where do we want AI and how do we want AI? We already know that these so called harms are actually unevenly distributed across society because data sets can be biased, as can the formulas through which we create actual machine prediction. And so that is the social phenomenon that I describe when I say AI social infrastructure. We sort of have in some way agreed that we want the logics of predictions to be the logics by which we organize society. And so that's something that has happened over the last couple years. And that's something that I'm interested in, in exploring further and also in softening a little bit and making a little bit more room for things like governance and participation, certainly.
[00:12:05] Renée: And when we think about it as a social infrastructure in itself, we're also thinking about the impact it's having on every social system that we are engaging with. So when we think about its long term impact on society, the ways it's changing social interaction, the ways in which it has sort of reimagined the social contract, and the ways in which we need to think about how we engage with the technology socially, what kind of advice would you have for anyone thinking of engaging with AI when it comes from just that social perspective?
[00:12:40] Mona: That is such an important question, Renée, and thank you for asking that. I think the answer to that is complex and evolving. As these systems are complex and evolving. I think it sort of starts with really understanding where the technology is at work and asking questions around that. That can be as you interact with your bank or as you interact with with your friends, or as you interact with system and colleagues in the workplace, trying to figure out where AI is actually at work and what is it actually doing in that specific context. Is it a summarization task? Is it actually a ranking system? Danny earlier spoke about the forthcoming session on recruiting. I've been working on AI recruiting for four to five years now, and AI actually plays a sort of more epistemic role where it ranks people or it assesses or predicts their future success and sort of plays a role in gatekeeping the labor market, if you will. So I think that is a really, really important component.
We also need to think more about what do we want from our fundamental social institutions and what role do they play. And of course, that is a big question we're addressing here at UVA as we are grappling with the infrastructural integration of AI into the classroom and education. I can give you an example from my classroom.
In order to keep our students focused on learning how to learn, which I think is one of the most important things we can do for our students, I have actually made my classroom low to no tech, so phones have to be in bags and laptops are research machines only and only come out for research tasks that are given in the classroom. And what I've observed is that students are a lot more focused and they are doing that sort of epistemic work themselves and they are happy about it. And it is a complete revelation, if you will. Going back to literally the drawing board, I'M also not the only faculty who's doing that.
And that comes from really asking, you know, stepping back and asking, what do we want from the, from the university, what do we want from the classroom? And asking that basic question vis a vis, a sort of nuanced understanding of what AI systems do. I allow them, for example, to use AI systems for research, specific systems, and we agree upon how they actually have to reference their use. So they have to, for example, when they use it in writing a research paper, they are allowed to use it for initial research and brainstorming. They have to give me the day on which they did that, which tool they used and which model of the tool and the prompt, for example. So we're like together exploring how we can be a little bit more literate about the sort of infrastructure and epistemic role that AI plays.
[00:15:41] Renée: Definitely. So I don't have the luxury of making it low tech because I'm teaching online. So that in itself is engaging with the technology. But I do ensure, and I always ask students that if they are using any kind of prompt, to share that prompt with me, to also share the process of the prompt to see how you began and how you end. Almost like a paper trail, right. Ensuring that you have, you know, your, your evidence that is there, which is so very important. I also ask them to engage with the technology because I do believe that everything we're using at this moment is infused with some measure of AI.
When we think about AI, there's a lot of fact, there is a lot of fiction, and unfortunately there's an extraordinary amount of fear. I know, particularly within the criminal justice system and policing, the technology has over promised and it has under delivered when it comes to sentencing and parole and the ways in which we're thinking about public safety. We know the technology has the ability to create these zombie predictions that overestimates the risks of black and brown and poor defendants in the criminal justice system. But we also know in health care and in architecture and in engineering, we're seeing the greatest potential of this technology.
I want you, of course, to, I'm not asking you to put on your magician hat or your psychic hat, but to tell us what do you think would be the fact behind this technology? What it can, what it cannot do. And of course, what are your future predictions about how this technology is going to reinvent itself over the course of the next five years?
[00:17:29] Mona: That is such a big question. And I guess if I really knew the answer, I would probably make money off that bet. So let me answer Sort of the first part first. So what can I do today? I think we actually all intimately know what I can do today because I think that most of us have at least played with the free generative AI systems that are out there on the market and have perhaps used them for things like drafting emails or speeches or probably for search, as you said, Renée, search, you know, online search is now generative AI powered as well. So I think what specifically generative AI can do today is it can understand the relationship between word fragments really well. So it's sort of a cartography of the human language, not all languages. I will also say English, of course, is overrepresented here very much. And there's a lot of research on underrepresented languages and machine learning that is out there and really interesting. So it does that well, which means it can help us with summarization tasks, it can create useful spreadsheets, it's good at that.
It cannot magically predict the future. Again, we're still in the realm of probabilities here, but it can work really well as a tool in very specific contexts. And these can include scientific processes. Right. And so we know that, for example, AI has already really excelled at drug development because the prediction of certain chemical behaviors can shorten the Runway to actual clinical trial with humans. And we all benefited from that actually as we were dealing with the pandemic and the vaccines were being developed. So it can work well as a tool and very specific and very well defined and very well structured environment.
Now, what is the future?
You know, just this past week we learned of Moltbook. I don't know, Moltbook. Sorry, I don't know if any of you read this. This is a social media page for completely autonomous AI agents that are talking to each other without human interference. It was sort of conceived of more as a little bit, you know, as an art project, but it has really sort of shown people where agentic AI is going. Which means that is an AI system that is granted access to our most personal text stacks. So our text messages, our email, our calendar and can sort of autonomously do things like creating calendar invites or making bookings. And we've sort of seen also that when it's sort of uncontrolled and without guardrails, it can get in sort of the more technical infrastructure and autonomously make changes. So I think agentic AI is where it is headed when it comes to language based models.
The AI research community isn't necessarily all in on. We've reached the end game. Generative AI through large language models. This is it.
We will achieve general artificial intelligence through that.
There are other researchers that say we need to think about world models, so we need to think about different kinds of data, visual data, sensory data, and create truly autonomous agents that have these sensory capabilities. And then there's of course quantum computing, which will absolutely change. I think everything that we're doing with AI because it's changing the technical infrastructure and the rate and volume at which we can actually compute.
So those are my responses to the future question, but thank you for the opportunity. Renée, what do you think? What do you think where it's headed?
[00:21:41] Renée: Definitely. I know we are in the age of agentic and I know the move is to superintelligence and we're trying to continuously amass all the data that we can to deliver these results. But yet we are seeing some critical challenges when it comes to the governance of the technology.
Because we're still seeing along the lines of gen AI, the questions around the technology really still challenged when it comes to the ways in which we are getting those responses that still speak to the bias, that still speaks to discrimination, threat, things like a geographical bias or the fact, as you mentioned, English as the primary interface and many other cultures and communities and languages not getting the kind of response from the technology. And of course questions around hallucinations still there, the technology doing things that it's not expected to do. So for me, I am really committed in the agentics, that space and really having a lot of conversations around superintelligence. And what is that supposed to look like from an ethical perspective and for a governance perspective, because we are still within the realm of governance, thinking about how do we regulate this technology. One of the things that we're seeing is that it requires real time governance and we know governance in real time are two things just don't happen together.
We're seeing the speed of the technology, the scale of the technology, the severity of the technology, and we're not seeing the kind of governance structures to match. When we think about the impact of this technology on children, when we think about the impact of this technology on the environment, we know that we need the technology. We know that the technology is working for us, but we also know the technology is working against us. I think for me it's always been conversations around duty of care. When we are building this technology, liability is critical. Responsibility is critical from the perspective of the individuals who are building it and from the perspective of the users and the people who are procuring the technology.
So having that level of collective responsibility as well as personal responsibility, as well as the question of duty to warn. I think the technology is still doing a lot of things that are quite concerning. And it is incumbent on the individuals and the companies, organizations who are developing, designing or deploying, warning people as to what the technology can do, what is fact and what is fiction, and putting those disclaimers in there so people could use the technology responsibly. We all know we didn't get a manual, we didn't get a playbook, and we're sort of doing it as it's being done to us. So it really requires a more robust and rigorous approach to governance. I always say to people, I believe that ethics and innovation could exist in the same space. We can innovate ethically, we can innovate safely, we could accelerate with this technology cautiously. But we must also understand that there are groups that need to be protected, particularly children, vulnerable groups, the environment.
These are all spaces where we're seeing a particular kind of real time devastation and harm that we need to pay more attention to. So for me, when I'm thinking, I know we're already in the age of agentic and more and more and more people, and of course companies are embracing it. But then again, there are some reports that show us that many companies that have tried to embrace this technology have not received the return on investment that they had hoped. And many companies are still not embracing the technology in a, you know, in, in that kind of real way. But definitely I see agentic on the future and I've been in a lot of conversations about superintelligence. So between the agentic and superintelligence space to see what can really be delivered, of course, always responsibly, ethically, and building trust in, you know, in the public space as well.
[00:25:53] Mona: Yeah, I couldn't agree more. And I think what we can really do is start in our own lives, right, and start in our own communities. And so one project that we are working on here at UVA through Zone Lab and in partnership with a vice Provost of Online Education and Digital Innovation, is to stand up a student Technology council here at UVA, which is going to be an elected and representative body of students who advises the administration and faculty on questions around digital systems, procurement, innovation and data governance. And so we really see that as absolutely fundamentally important because technology is power, bureaucracy is power, and students have been sort of more seen as consumers of digital systems rather than actual stakeholders. In the age of AI, that sort of no longer holds up because we are creating data as Renée said earlier and we are creating digital footprints that end up in predictions that can have an impact on our lives in five years, 10 years, 15 years. So the data footprints that we're leaving and that our students are leaving as part of being enrolled at UVA have a, you know, have a potential impact on their lives and so they ought to have a say in how this happens. And of course here at UVA we take student self-governance and governance and leadership quite seriously. And we are working on this and hopefully going to see a student technology council in the future. I think that that could be a good model for how we maybe want to think about community and governance and power in the age of AI. But Renée, let's stick with governance just a little bit more. Just because you are so tapped into the global, global debate and the global policy discourse.
For some it might seem that Pandora's box is already open, right? The technology is out there. We have Moltbook, we have agentic systems that out there, AI's infrastructure, which means we don't necessarily see it until it breaks and causes harm.
So from a policy and maybe global policy standpoint, where do you see real opportunities for ethical AI use? And I will also say productive aius. This is not just about doing the right thing. This is about making sure that these systems actually are useful, actually increase productivity and making sure that we're not living in a big bubble that may burst and have severe cascading effects on the global economy. So what are some of the, what are some of the most meaningful policy conversations you are having when you talk, when you go to New York and you speak at UN level or another global AI policy conversations?
[00:29:00] Renée: Mona, I think you touched upon it just now. And I always say the global begins with the local. And it's so important to have those kinds of stakeholder conversations, student led conversations, community led conversations, our state level conversations are conversations that also build what that global perspective is going to be. You know, I'm committed to public interest technology and ensuring that the public interest is always center to this technology and public values are centered to that technology. And I think the world is seeing that, seeing the importance of having national AI strategies, national data governance strategies.
So countries and states and communities have to be thinking about their data as well. It should not only be a global conversation. I think we also need to understand this question of governance. And as you said many times, students see themselves as users. And one of the things that I always say to people that we are producers of this data that's being scraped and re harvested and repackaged and resold and, and re everything. And we have to understand the power of that data. Yes, we are sort of coming from behind the eighth ball because we never knew that data was going to be something so powerful. We never knew that this thing that we can't really see, feel or touch was going to be the sort of greatest geopolitical game changer in the history of the world. We never thought that AI was going to really impact geopolitics in the ways that it's impacting geopolitics at the moment. We never thought that it would be a question of political power and even a question of energy. When we think about the power that's required to ensure that we get the compute power that we want. So at one end it's geopolitical power, it's international power, it's political power. But when we think about the impact on the environment, we're thinking about energy, we're thinking about water, we're thinking about that kind of power as well. So on the global level, the conversations are of course looking at the ways in which the future is now being impacted by the technology, looking at the kind of power this technology has and looking at the divides that are being created between emerging markets like the Global south and established markets like the Global north and where that concentration of power is and how you can have tech companies now negotiating at the highest level with countries and world leaders and how those relationships between the big tech and the governments are impacting our relationships in the ways in which we're using the technologies, which technologies we're using, the companies that are getting more leverage to innovate and more slack to innovate as well. And whether or not we're really standing up the kind of compliance and enforcement and, and sort of responsible AI approaches that are required, or whether or not we can get a handle of what's happening. Because we know that the law continues to lag behind the technology.
And we know that most governments have not really stood up the kind of legislation that's required. And we know even on a national level, many of the states are doing great things. But we know we have no real national policy when it comes to data. When it comes to AI. We know we've been very flexible in the US because we want to see where this technology goes or where innovation takes us. So these are really big questions. And the challenge is this. Lots of questions, not enough answers and not enough solutions and just the conversations continuing. But this being really a right time to get involved in the conversation, because I keep saying it, what happens on a personal level, we both know the personal has always been political. And what happens on a local level is also critical to what's happening on a global level.
[00:32:59] Mona: Thank you for that, Renée. So I, I wonder how you see sort of narratives around extraordinary technical capability versus what we know about how humans actually operate.
Right. And sort of the, the promises and the successes, but also the failures of societal organization and societal ideas and political ideas.
What do you feel is the sort of bigger threat and bigger opportunity, what the technology can potentially do or what the ideas that people put out there in the world and enact?
[00:33:40] Renée: So definitely an extraordinary amount of hype around what the technology could do. But. But then again, if we're looking in critical spaces and places, let's say healthcare, cancer research, we're seeing some really brilliant things with this technology. And just about, I would think every business model is sort of being reimagined and reinvented. But we always talk about AI as this big thing, and I think I always try to bring it back to data. There is no AI without data. Data is the lifeblood of this technology. Our AI is only as good as. As our data. And if our data continues to be challenged or sometimes compromised or traumatized, then we're building systems that are going to reflect the data that we have. So whatever the building is or whatever the reimagining is, we've got to always look at the data that we're using to create, because it's the data that really either delivers or under delivers. But I'm going to flip the script and I'm going to take it to you. And one of the things that we have to talk about is this question of human risk versus machine risk. I know in 2016, ProPublica this report that looked at machine bias in the criminal justice system, and it was really a point of reference for so many of us working in this space thinking about algorithms. But I have a quote that I am going to share with you, and the quote is, is I'm not afraid of artificial intelligence. I'm afraid of natural stupidity. And it's attributed to the Guillermo del Toro. And I'm going to ask you now, do you agree or do you disagree? And are the bigger risks tied to the people building the technology, the people using the technology, or the systems around the technology?
[00:35:34] Mona: So, as a sociologist, I'm always interested in people and people first and societal structures first. And I look at how these actually get expressed in technological systems and arrangements, including AI.
So my interest definitely is with people. I am not worried or afraid of human stupidity. I don't think that that is the right word. I think that that actually is a disempowering narrative that we are too stupid to actually understand what's going on, that we all are non experts in AI systems and therefore are not deserving of participation in a public and policy and innovation discourse around artificial intelligence.
I'm rejecting the formulation here a little bit. What I will say though, what is new about AI is that it really walks in the footsteps of bureaucratic decision making. It's a bureaucratic technology because it is automating decision making and it is designed to do more with less. It is a scaling instrument.
If there is a problem in the way in which that scaling happens, because data is reflecting societal biases or because we are or articulating an objective function that optimizes for things that can end up being harmful, such as more clicks on social media, which then heightens certain content that can be addictive, for example.
So we sort of need to be aware of that scaling function. And so I think in that sense it is human and the particular design of the system together that we need to keep in mind as we articulate where we want to go together as a society, where we want to go as an institution, how we make space for innovation and creativity and how we make sure that this actually happens with sustainability in mind so that we can have longevity and creativity and research and prosperity. I think this is not about stupidity. This is about peeling back what actually is at work here. And I do think that is it is absolutely essential that we listen to folks from the humanities and social scientists and actual communities who live with these systems in order to do this well.
So I have a job question for you, Renée. You know, a light one.
What kind of advice do you give folks who are looking for jobs who are encountering an AI mediated labor market, who sort of are acutely faced with promises of AI, who are facing demands on themselves with regards to rapidly evolving skill sets?
What would you say to folks and alumni who are interacting with recruiting systems that are AI mediated?
How should they deal with that? And how should our future graduates think about AI?
[00:38:56] Renée: Definitely. Well, I think one of the things I often hear first is the challenge of actually being hired. And you specialize in that area how these recruiting systems are impacting individuals. I have lots of friends who are looking for jobs. Some of them are thinking it's ageism. Many of them say these video interviews or these AI infused video interviews are working against Them. We know individuals with disabilities have spoken about how these technologies could really undermine your ability to get a job. So there's that challenge. And maybe after I finish the question, you can just address that. How these recruiting tools are really impacting and undermining people's possibilities and potential may not be the selling point at that moment. But really, how do you trick the algorithm to get beyond the recruitment process? But when it comes to actually working in an augmented AI space or an AI workforce, I really speak always about that sort of interdisciplinary collaboration and interdisciplinary thinking that is so required. Every industry, every business model, every approach is being impacted by this technology.
I will say think always about that ethical approach. And what are the ethics that are required? Of course, critical thinking, so important to the kind of judgment that you are bringing to the technology creativity. I think what we're seeing right now is the best of human intelligence and the best of machine intelligence coming together to stretch the imagination of imagination to make us more creative and domain expertise. That is so required. But I think just before we go into the sort of rapid fire and the final part with the questions, it would be important for you just to share about how the, you know, the recruitment process has now been really impacted by AI and this concept of sort of, you know, tricking the algorithm to get a job. You're hearing more and more about that right now.
[00:41:02] Mona: Yeah, the world of recruiting and human resource management has definitely changed since I began this research in 21.
It's important to note that recruiting in HR is never a business or a profit center within corporations or organizations. It is always a cost center.
Even recruiting agencies have who sort of staff and recruit professionally have very small margins. And so they're very careful about what large technology licenses they invest into.
So that's important to keep in mind. Right. These are. This is a profession that's already dealing with constrained resources. There are two types of recruiting generally. One is high volume recruiting, one is low volume recruiting. High volume recruiting is a situation where there is one sort of standard role that a lot of people need to get hired into.
That can be service work, but they can also be entry level jobs, for example, for our graduates or internships. And the idea here is to use technology and a bureaucratic system like AI to reduce the big, big pile rapidly and get to the ones that recruiters either hire directly or want to screen in a personal encounter, whether that's zoom or in person. The other end of the spectrum is low volume recruiting where the sort of talent market is really scarce. These are folks who have very specialized skill sets and where companies and organizations are competing for folks who are in that small, small pool that can be nurses or electricians or they can be people who have PhDs in fair machine learning where there are not a lot of, of folks around yet. And so here they use systems to get to identify candidates so as a search in a, in a search function, but also to make it a lot more likely that a candidate responds to their message when they cold reach out to them.
And a company like LinkedIn has really perfected that through their sort of really somewhat monopolistic platform and the InMail system and the internal tools that they provide to recruiters for a fee. So that's another important thing to keep in mind, right? Is are, are you in sort of a high volume space or are you more in a low volume space? And the use of AI really depends on that and the, the space that is given to recruiters in terms of discretionary decision making in the high volume space, they don't have a lot of discretionary space. They really have to make sure that they get people hired quickly and repeatedly. Forever. Green roles, for example, the discretionary space is much bigger on the low volume recruiting side where tacit knowledge and gut feeling still play a role.
The dynamic that we're seeing a little bit is that recruiters really want to hold on to their discretionary decision space and their tacit knowledge and the authority that comes with it, which means they're actually sometimes going low tech. And so interpersonal networks relationships are actually becoming more important, not less important. They often were taught make yourself legible to the machine and you will get through. That I don't think is a viable rule of thumb.
I think curating a presence that is legible in a way online that we want it to be legible is important. But really trying to tap into interpersonal networks and having an interpersonal skill set is going to be the sort of skill of the future.
Now what we're seeing is that a lot of folks who may have been in the low volume recruiting space are moving over to high volume because there are more facing higher rates of potential AI automatability. And that is something I think we need to watch.
Whilst I cannot give a very clear rule of thumb and I wouldn't necessarily subscribe to make yourself as legible as possible to a machine, I do think that there is something to be learned about curating oneself's professional identity online and, and tapping into personal networks if and when available.
[00:45:47] Renée: And I'll just ask you my final question, which industry do you think will be most disrupted by AI? If there's one or many?
[00:45:57] Mona: Advertising and marketing, I think and sort of more on the high volume end of things. I think that as a, you know, the dynamic we see everywhere luxury is human curation we will see in marketing and advertising, but we're already seeing a very, very high levels of automation in that domain. So I think that that is something that is happening and sort of language based tasks and highly structured environments which also is already happening.
So Renée, what is one thing you'd ban AI from doing?
[00:46:39] Renée: For me, anything that is a life changing decision.
So sentencing, parole, juvenile justice, anything that really deals with due process, very concern.
And I would not ban it, but I would put up some very heavy protections around therapy and any kind of engagement in the mental illness space.
[00:47:06] Margaux: Thanks for listening to this episode of UVA Data Points.
More information can be found @datascience virginia.edu. if you're enjoying UVA Data Points, be sure to give us a rating and review wherever you listen to podcasts. We'll be back soon with another conversation about the ever-evolving world of data science.
