Season 8 Episode 3: Ethics, Engineering and AI
Season 8 Episode 3 [download this episode (MP3, 23.8 MB)]
In this episode Elizabeth chats with Jason Millar, a professor in the School of Engineering, Design and Teaching Innovation in the Faculty of Engineering at the University of Ottawa, and Canada Research Chair in the Ethical Engineering of Robotics and Artificial Intelligence, to unpack ethical engineering in AI. They discuss why concepts like safety mean different things to ethicists and engineers, and how that translation leaves harm like misinformation and emotional dependence on chatbots unaddressed. Jason shares his work at CRAiEDL, and why incentives, regulation, and engineering culture matter. He argues policy makers shouldn’t assume engineers will get this right with regulation.
Additional Resources:
Check out Jason’s academic work on Google Scholar!
Read Canada’s proposed The Artificial Intelligence and Data Act (AIDA) which was part of Bill C-27 that was halted at the beginning of 2025.
Tune in to Christelle’s episode on the governance of AI
To learn more about the history of engineering and how we have seatbelts in cars read Jason’s early work An ethical dilemma: When robot cars must kill, who should pick the victim?
You can find Jason on LinkedIn as well!
Contributors:
Host: Elizabeth Dubois
Research Lead: Areej El-Sharif
Transcript: Areej El-Sharif
Audio Editing: Sarah Pledge-Dickson
Translation: Julianne Bernier Robert
Publishing and Promotion: Aaron Brisson
Producer: Sarah Pledge-Dickson
Episode Transcript: Season 8 Episode 3: Ethics, Engineering and AI
Read the transcript below or download a copy in the language of your choice:
Elizabeth Dubois: [00:00:04] Welcome to Wonks and War Rooms, where political communication theory meets on the ground strategy. I'm your host, Elizabeth Dubois. I'm an associate professor and university research chair in politics, communication and technology at the University of Ottawa. My pronouns are she/her. This season, we're talking about social and policy issues surrounding uses and governance of artificial intelligence. And today we're talking with Jason Miller about ethical engineering. Jason, can you introduce yourself, please?
Jason Millar: [00:00:27] I'm Jason Millar, I'm a professor in the School of Engineering, Design and Teaching Innovation in the Faculty of Engineering at the University of Ottawa. And I have a Canada Research Chair in the Ethical Engineering of Robotics and Artificial Intelligence. It's a mouthful, but what that means is basically, I spend most of my time researching the ethical implications of robotics and artificial intelligence, and thinking about how we can change engineering practice and policy to respond to the kind of issues, ethical issues that those technologies are raising today. And there are many.
Elizabeth Dubois: [00:01:05] Yeah, there are many. And so I'm really grateful that you're able to take some time to chat with us. The idea of ethical engineering ethics in AI is obviously something that's super top of mind for a lot of people. One of the things that I like to do at the beginning of the episodes this season is do a little level setting, because I think when we're talking about AI and insert any additional concept, we often are talking about a lot of different things. We're not necessarily all thinking about the same set of problems. So can you give me kind of the short landscape of what ethical AI or ethics in engineering in this context means? What are you thinking of when I say that?
Jason Millar: [00:01:44] Yeah. So I've been studying ethics as a philosopher first for years. So this is going to be my slightly long winded answer. You know, I so coming from a philosophy background and I have a dual background in engineering, which is how I ended up in an engineering faculty studying ethics, you know, for years as a philosopher and really getting interested, you know, maybe 15 years ago in topics that ethicists were talking about that had to do with technology. So at the time, you know, years ago, robotics, AI just kind of like, you know, interesting topics. They weren't widespread in society. We weren't seeing, you know, videos of robot Olympics from China every day or interacting with LLMs back then, obviously. So we, we barely had Siri back then, I don't think, but as a philosopher and an ethicist, what I realized when I was speaking to engineers which naturally just happened through the work, the kind of research that I was doing as a grad student, I started to realize very quickly that if you walked into a room filled with engineers or even policy makers, and you were sort of speaking the language of philosophical ethics, as interesting and fascinating as that was to me and other philosophers who do that kind of work. You may get people in the room interested in what you're talking about, but the engineers, for very practical and understandable reasons had no idea, for the most part, how to integrate that into the day to day work that they do. So it's, you know, you were asking me like, how does ethics and engineering work for me? I've spent a lot of time in my career and my research, and certainly the work that we do at the lab at the University of Ottawa try to find ways to bring that philosophical conversation into a much closer contact with engineering practice, to develop the kind of language and tools and practices that engineers can use to really bring those ethical considerations more centrally into the work that they're doing when they're designing robotics or artificial intelligence technologies.
Elizabeth Dubois: [00:03:52] That's really helpful. And so that kind of bringing it in, I've heard you talk about it as a translation and there being translation problems, right? So it's not just about, oh, well, we need to use the language that the other group is using, but also the implementation of it becomes a challenge. Can you talk a little more about that idea of it being a translation problem?
Jason Millar: [00:04:16] Yeah, absolutely. So, you know, well ethicists, we talk a lot about ethical principles. So you might have something like, you know, privacy or even like autonomy or something like that. Like these are words that philosophers use to that are very packed with meaning, right? They mean a lot in terms of what individuals, what their responsibilities are or their duties to one another. Or if you want to frame it that way, sometimes, you know, how we should structure relationships or, you know, think about technology in terms of like, what's the technology doing to people? What shouldn't it be doing? Like, how can we, I mean, we have these languages like that. So autonomy, even safety as a word means something very particular to ethicists and philosophers. And this is true. I mean, what I'm describing here, I'm about to describe is very true of any group of people who study something or who have like a working, you know, a kind of a community of practice. Lawyers have this policy makers have this, sociologists have this, historians have this and engineers and philosophers also have these languages that they use, right? A lot of times we share the words. A word like safety seems like it might mean exactly the same thing to an engineer as it does to a philosopher. But when you get in a room full of philosophers and talk about safety, say AI safety, and then you get into a room filled with engineers or policy makers for that matter, and talk about safety.
Jason Millar: [00:05:39] What you start to realize is that most often they're talking about similar things, but quite different in the implementation. So, you know, safety for somebody in the humanities, philosophy, ethics, sociology may mean like caring about the well-being of an individual that could include like their mental well-being, their physical well-being you know, and that could translate into all sorts of conversations that you have about helping them maintain healthy relationships, whatever it is. I mean, these could be concepts or ideas that fall under that umbrella of safety. You go into an engineering context and, you know, talk to people who are doing quote unquote, safety engineering, which has a long history in the profession. And there it's only indirectly related to the kinds of things that the philosophers might be talking about in an engineering context. Safety engineering quite often will mean, let's say you're developing a car and you're doing the safety engineering for the vehicle. Well, then there are going to be a bunch of tests that you have to run on different subsystems, the braking subsystem, the steering subsystem to make sure that it's designed and it works in a way that is very predictable and passes these tests. So for example, you'll have to design the system so that if you put a certain amount of force on the brake pedal. It will be able to slow the vehicle from a certain speed to a certain speed in a certain distance.
Jason Millar: [00:07:02] So very mathematical, very objective, very kind of lab oriented, right. Numbers and forces and accelerations and all that kind of stuff. But you're not really talking about what that means in terms of keeping a person happy or, or physically intact. It's only very indirectly related. It's related, no doubt, but the concepts that sound the same. So if you're in a room and a philosopher says, we need safer technology, engineers may be hearing something that's very different from what the philosopher actually meant, because they may be referring to these other things. So there's a translation that happens from the word safety into the testing regimen that the engineers have to undergo. And that takes a lot of work, right? It takes a lot of negotiation with regulators, with standards bodies, with companies. And, and that's the kind of translation work that I'm interested in, is trying to figure out how to go from these concepts that we have on the one side into working engineering practice on the other. So that's the translation piece in the bit, kind of quantifying it in a way, and making sure that it meaningfully reflects the intentions that, you know, the average person would have. When they say, hey, I want to know whether that vehicle is safe or not. Well, now we want to know whether this AI is safe or not. There's a lot of work that still needs to be done to get there, right?
Elizabeth Dubois: [00:08:20] As a methods Prof. It reminds me of going from conceptualization of some variable to operationalization of some variable, right? Like you want to measure how safe a thing is, safety could be a whole bunch of different things. First, you need to hone in on conceptually what you mean by safety. And then you have to go even more nuanced to a narrow definition of exactly like this is how we will measure whether the thing is safe or not. Here are the very specific standards we're looking for.
Jason Millar: [00:08:49] That's right.
Elizabeth Dubois: [00:08:50] And so like, that's a really, really complex and challenging Process. And when we're thinking about the kinds of harms that we have been forced to think about with AI, things like huge environmental impacts. Yes, some physical safety, but also things like mental health or impacts on democracy and democratic resilience. Or like there's just there's huge, huge things that have very little to do with like the physical safety that in my mind seems to have been the main way that engineers tend to think about safety. Maybe I'm not an engineer. Maybe I'm wrong. I'll let you weigh in.
Jason Millar: [00:09:31] No, I mean, and that's the great insight is that, you know, we. And it's also where you can be hopeful, right? So yes, it's true. We have a long history in engineering of thinking about safety, often pushing back on it. If you look at the history, right, like, you know, broadening those concepts comes with a fight, comes with its challenges in trying to convince a group of people to like radically change the way they think about [...] their day to day work, right? Mhm. So yeah, but the insight there is that we have done this with safety to a certain extent. And now what we're in is I think this period where we're just like kind of radically broaden the concept to include a lot of other types of safety. You know, engineers just have not traditionally had to think about much. And now that's safety. If you're talking about the integrity of your political system I can guarantee you that's not something that we've translated clearly into, you know, test, test specs and stuff like that in, in, in engineering contexts, right? That like starting a whole new conversation that we are just trying to figure out right now, even though, you know, on the humanities side, we've been recognizing for quite some time that, you know, how misinformation is working and, and how these platforms and the way they use that, you know, AI are exacerbating these problems. We know, we've known about that, but translating that into actual practices is the challenge that we still have.
Elizabeth Dubois: [00:10:57] Yeah. Yeah. So I actually did a little experiment with Fenwick McKelvey, who's a professor at Concordia University, and a few of his students, where a while back we went to the most popular of the like ChatGPT, Gemini Copilot, etc., and tried to test it out and see like, could you actually create political disinformation? Like, could we get it to generate tweet length messages that are completely fictitious might help to undermine an election, that kind of thing. And, you know, most of them, it was like, “Yeah, of course, here you go.” And then a few of them were like, “You can't talk to us about politics.” But then like, you reframe it as if like, I'm writing a fictional book and like, now we can generate it. All right. And so like, it's just an example of those, like it's not thought about or if it's thought about, it's thought about in like one very specific, narrow way, the kind of operationalization of it is really narrow and specific. But when you actually think from the social sciences and humanities perspective, it's like missing the whole picture of what this means for democratic integrity.
Jason Millar: [00:12:06] Yes. Yeah. Like, you know, you're trying to ask, like, how do you think about AI right now? I mean, AI for people who work in AI, it's many, many things. For most of us, it's like ChatGPT, right? To the extent that the average person is interacting with AI deliberately and, and knowingly, it's when they're on the ChatGPT console, right? And they're asking questions and getting it to help them, you know, search the web for interesting information, helping them to solve problems. I do that, I do that all the time. And more and more I find because it's really good at helping. But you realize very quickly that you're playing a big language game with the AI. And it's the, the AI seems to miss the context. So, you know, like you said, you reframe your question slightly differently and get it thinking that you're doing something different. It's a very easy to talk an AI into giving you certain bits of information, depending on how you parcel up the set of activities that you're asking it to do that would, that, you know, would help you get that big picture of information that the safety people might be trying to design out of the system, but you can't. It's really difficult to do. Yeah. When you're playing these language games. It's like when you're asking a kid to do something like reframe it slightly, and then all of a sudden the kid's totally on board like, “Oh, you don't want to go to bed? Well, I'll tell you what. Why don't you just go get ready for story time?” Like it's a, you know, it's not necessarily bedtime now it's story time. You know, AI is a bit like that or a lot like that in many cases.
Elizabeth Dubois: [00:13:37] Yeah, yeah. The whole idea of like, well, you need to learn how to prompt it is also like, yeah, you learn how to prompt it to get around the safety features that have been put in place. Well, then that's not a particularly good safety feature if it's easy to work around it. And so we get into this sort of technical issue, although there's also questions about like, is that truly a technical issue? Like, is that actually a necessary limit for these platforms? Or like, if engineers believed it was their job to fix it in a different way, or if there was better translation the way you're describing, like, I guess what I'm getting to is how do we solve this problem, right? Like right now, there's been some initial attention paid to particularly generative AI tools and chatbot kind of tools, conversational agents. Okay. There are these harms. They are societal harms and they are nonphysical harms that are not necessarily thought of right now or built into existing engineering safety practices or responsible AI practices. Like what do we –
Jason Millar: [00:14:43] Like people forming deep emotional relationships with chatbots, right?
Elizabeth Dubois: [00:14:47] Yeah.
Jason Millar: [00:14:48] You know, there's no course in that, like in an engineering or computer science faculty, right? Or at least there weren't four years ago or five years ago when they were building these things for the first time.
Elizabeth Dubois: [00:14:57] So what do we do?
Jason Millar: [00:15:00] Yeah. I mean, you know, one of the things that's interesting about AI and possibly robotics, [...] you know, when we see these images of like – I watched a video and I'm, you know, if you get on Instagram and just look around for the video of the robot kicking some little kid, one of these humanoid robots that's doing some sort of a, like a kung fu demonstration. Like, this is what I can the best I can gather, this is what it was doing. And it just kind of, you know, the technology is at a point where it's not quite able to really totally understand where it is, what the objects around it are, and form like a set of internal instructions to like really protect those, those objects around it. Like these things will kick TVs, they'll kick a chair, they'll kick whatever accidentally, including little kids like. And you see these videos of little kids getting hurt by these things. You know, similar to the AI, we're at a point right now where what's interesting is that we're releasing these things into the public in these social settings, and just performing these massive experiments without even having those types of safety checks in place. Right? There's really no world where it should be acceptable to put a robot in a room with kids. And there's a real possibility that one of those kids is going to get physically hurt. And yet there's the video, you know, in our lab, the Canadian Robotics and Artificial Intelligence Ethical Design Lab. It's CRAiEDL. When we're doing our work, there are three overlapping areas that we just feel are, you know, always entangled with one another.
Jason Millar: [00:16:39] So it's the, you know, the underlying ethical issues that we're trying to understand, like what kind of ethical issues do we need to deal with with these technologies? What do they raise? What is the kind of engineering practice that we can adopt to help designers of the systems avoid or manage those ethical issues, because there are ways. And then what's the kind of policy environment that we're working with here, and how much heavy lifting do the engineers have to do versus because of, in many cases, the lack of policy or regulation to do that work for them? So right now, we're dealing with a regulatory environment where there are no real safeguards for these types of issues that we're talking about. The physical thing. Yes. Like if, you know, in North America, somebody put a robot in a room and it kicked a child, there would be lawsuits and, and, you know, possibly even engineers losing their licenses to practice and being fined or even worse, going to prison. Like we have some of these safeguards in place. And yet, [laughs] so, you know, we try to understand like, what's the regulatory environment that needs to be brought into place to, to eliminate some of these very basic things. And there's a possibility that some of these technologies, the engineers, don't have a real way of solving these problems because they're complex problems. How do you get an LLM to understand the language context that it's working within? And to understand that you or I are maybe trying to fool it into giving us information by reframing questions and things like that.
Jason Millar: [00:18:17] If it's the case that the engineers can't put safeguards in place to protect against that kind of behavior, then it raises real questions about whether, like, are the benefits so outweighing the risks at this point that we're just willing to allow that in a society? I mean, I, you know, I have questions about this. That's part of this is the work that we do. There are certainly a lot of amazing things that AI can do. I'm not sure you need a publicly accessible LLMS with all of this information built into them in order to get the benefits that you would otherwise get. I think people are starting to realize that we have, we've really missed the goalposts here when it comes to providing an AI environment that is responsible for widespread distribution and use. And it's, but it's a serious problem. One of the interesting things is that the question is, well, then what do you do? How do you regulate. It's not clear to me how, you know, if you follow the story more and more. What I'm finding interesting is, you know, when I hear podcasts or interviews with some of the leaders and people who are really spending all of their time trying to figure out how to make AI safe, they're at a bit of a loss for words. And that's something that I find worrying suggests to me that we've gone down a path that's maybe leading to some bad things.
Elizabeth Dubois: [00:19:37] Yeah. It seems to me like we just accepted, took for granted that the choices were either AI is on or AI is off. And by the time we were confronted with that choice, it felt like the horse was already gone and out. And so it was like, well, we can't rein it all in. So I guess we're going to have to try and be in the AI game. This is government saying we need to catch up. We need to be part of this. This is where growth is going. And so all of this money and excitement has been poured into it. But we didn't pause to think, “Okay, maybe there's a different shape for the AI to take.” And like the way that this should be rolled out could be limited. Like what you're saying there about, you know, these LLMs don't have to be able to do everything right. We could constrain them, we could contextualize them. There could be boundaries set up when there are instances where the solution to whatever safety problem is there or potential harm is there, like we just don't have that yet. So maybe don't deploy that component of it until you do have that, you know, I mean, is it too late to have those conversations? Where do we go when we're now in this stage where there aren't great answers to some of these. Minimize harms and prioritize safety questions.
Jason Millar: [00:20:54] Yeah. I mean, you know, the technology is so, so accessible now, the knowledge to build the technology is so accessible and the costs are relatively low. I mean, yes, if you're trying to build one of these massive data centers, yeah, it's going to cost you a tremendous amount of money. But there are even the smaller, you know, open source models that are available everywhere now. There are just so many of them coming out of China, even coming out of, you know, meta. And if you go to Hugging Face, you can find these things like they're out there. It's really not clear that we can put things back into the bag now. I mean, sure, yes, we can. Like, so theoretically we're in control the right now anyway, that the eyes are clearly not in control. But, you know, we spent years and years early on in the AI debate. There were these kind of two camps of AI ethics, responsible AI people. This is like just speaking very broadly, there were the people who were very worried about existential risk. “Oh, the AIS are going to destroy everything. They're going to like do serious damage to humanity. And like, we need to start studying existential risk problems” versus the people who and I fell into this latter camp were more like, “Hey, can we just make sure that we have off switches on these things so that whatever system you put in place, they're not just spinning off and doing things without any meaningful human control in there, without any really meaningful oversight. And if we can't get the oversight done right, then we [...] need to figure out how to do the small things like how to protect against like the little harms, the misinformation, all that kind of stuff. Let's clamp down on that.”
Jason Millar: [00:22:38] But, you know, the existential risk is kind of a sexy problem. It's the Terminator problem, right? So from a policy perspective and from a research funding perspective, there were a lot of resources poured into that. Challenge there is that you're always thinking of this future hypothetical. You're never really focused on the risks and damages you're doing today. And if you know, if you could fix the little things that are happening today, which are going to seem small compared to those big things, you're probably going to prevent the big things as well. That's the working hypothesis anyway. But unfortunately, we spend a lot of time thinking about these future scenarios that, you know, where there's an and it's like, “Hey, well, what if we had this kind of system? Yeah, that'd be bad. We can't do that.”
Jason Millar: [00:23:22] Meanwhile, we're just doing all of that here. So this is my long way of saying that, you know, are we past the point of no return when you take your eye off the ball for so long and you're focused on these big things, like the catastrophic thing that's coming. You miss the little, the million little things that are piling up. And then all of a sudden you realize, “Whoa, we don't know what to do with it.” There are so many of these models out there, so many different countries developing them, so many resources that are accessible to people who want to use them. How do we now, you know, go back ten years and we can't like so it's it's actually the, you know, question mark. Are we past the point of no return? It's not clear to me what kind of political will would need to happen short of real international coordination on regulation. And there's just I don't see that happening right now. I'm not trying to be a pessimist. I'm just the reality sinks in when you look around and you're asking like, well, what are the practical solutions? The honest answer is that it's not clear what they are. You know, we can do things in Canada that won't stop it at this point.
Elizabeth Dubois: [00:24:32] Yeah. There's a whole geopolitical context here too, that all of this plays into, which we don't have time to chat about today. But in some other episodes this season, we'll get into that a little bit more. I want to go back to the idea of the translation and the standard setting, because I think that starts to play into these smaller things, whether or not doing a bunch of them, you know, 15 years on when we haven't done much of them yet, is going to be the thing that turns the tides. We do still know that there are those kind of practical level things that need to be addressed and dealt with. So when we're thinking about the standard setting, who gets to do that? Like, is it the engineers themselves coming up with these ideas? In that case, that feels a little weird to be like, “Hey, engineers who are being paid by a company that's trying to make lots of money off of this, figure out what safety standards you need to meet and then go meet them.” But they're also the people who maybe have the most technical knowledge to understand what standards would be appropriate, but they're also the ones who are furthest away from like the sociotechnical scholars who understand the bigger problems, like what work, what roles goes into creating those standards beyond the like political will for government to actually enforce them.
Jason Millar: [00:25:45] Yeah. Well, so here's, I mean, like where I'll put my optimist hat on and, and that's that regardless of whether we're headed for, you know, some terrible event that forces us to come around as a community like international community and start doing things differently. We, there's always the opportunity to do things differently, right? So [...] you may end up just suffering a little more along the way to getting it right or getting it done better. Right. So I'm an optimist in that sense. There are solutions to these problems. We just, we need the incentives to get that done. So [...] this is where things get complicated again. But this is the reality that we're working within. So one of the approaches is to, you know, slowly over time, but [...] very deliberately and intentionally chang[e] the way that engineers understand their responsibility. Engineers and computer science technologists in general understand their responsibility vis a vis technology. Right? When I'm designing something as an engineer, what is it that I really need to be taking into account? Just because I've been told and socialized to believe that that's the right way to do it, right? We believe, you know, most engineers graduate now with an understanding that they need to be responsive to environmental concerns, to safety concerns. And now we're talking about broadening that. So that's one aspect of it. What we want to prevent is those engineers graduating and going off into companies where the regulatory environment is such that they are immediately told that none of that stuff matters.
Jason Millar: [00:27:15] “Yeah, yeah. You learned about that nonsense in university. Here's how we do things here, and you will be rewarded greatly for doing them this way that I'm about to explain to you.” Right. That's what we want to prevent. And then they go off and, and just, you know, disregard all of the things that they've learned about how to design responsibly. So we're talking about incentivizing at a different level there, right? Corporate incentives. So that's much more complicated. Standard setting is one way of doing that, but standards are voluntary. Corporations have to come together and decide that they want those standards. In a world where you're constantly. And I've sat in these rooms with these companies. You know, I worked for an organization called Partnership on AI. I was the director of research there for a couple of years. That's a large collection of the biggest companies Amazon, Apple, Google, and so on. All the big companies are in there and they talk about these issues. They talk about doing it better. They talk about trying to figure out how to design AI so that it's not doing all these terrible things. And what you realize very quickly is that they are dealing with an incredibly competitive marketplace where if there are no government regulations, laws, things like that to prevent them from doing things, then it's a race to the bottom.
Jason Millar: [00:28:29] And that's exactly where we're at right now. We're in an environment where the incentives don't exist properly. The incentives are not designed to prevent the kinds of things that are happening. The incentives are not designed to motivate the development and uptake of regulation to prevent certain things. I mean, look, right after ChatGPT first came out a few years ago, a lot of these executives for these companies signed a letter saying like, “Hey, we need a six month pause. This is crazy.” That was three years ago, right? And the regulatory environment hasn't shifted at all in that intervening time. So the incentive structure has not shifted at all. In fact, it's become more motivated to do this race because of the amount of money that we're seeing in this game. Now. The stakes are so high. Now, if it's not existential for us yet, it's definitely an existential threat for the companies who are investing hundreds of millions of dollars every year to try to build data centers and grow models and things like that, because the way the future looks is that from their perspective, this could all go away very quickly if they don't beat the other companies. And in a regulatory environment that's basically non-existent, that means doing whatever you want to beat those other companies.
Elizabeth Dubois: [00:29:45] Yeah. Which does not bode well for the idea of let's like slow down and really deeply think about the potential causes of harm that we might be creating in these tools.
Jason Millar: [00:29:57] Which is what philosophers like to do. Slow thinking, take your time, get it right. Slow down a little bit. Yeah. No. So that even that's like a creates friction between ethicists and, and engineers, where the time horizon that you've got to solve these problems. It's like, we need these answers yesterday. If you don't have them, then we're doing whatever we need to do to get this stuff out the door.
Elizabeth Dubois: [00:30:18] So what I'm hearing from this conversation is we need a set of tools, whether that's particular standards or particular measures. What exactly that looks like is really up for debate and needs a lot of thought, probably from a variety of different people trying to figure out what does it look like to actually create the kind of safety we want. But then we also need this kind of governance structure, this enforcement process, to make it happen, because the current setup is not incentivizing the creation of any of that. And then we also need this culture among engineers of like, “Actually, it is our responsibility to think about this and implement this,” and you need buy in from the people who are creating these tools because regardless of whether the tools exist and whether the governance structure exists, really the people who are going to implement it need to be, “Yeah, this is my job and I care about this.” Right.
Jason Millar: [00:31:12] That's right. The goal here is to create as many reasons and incentives for people to do the right thing as is humanly possible. And we are not doing what's humanly possible right now. But we're but we're talking about it and we are developing those tools. So we're developing methods that we are using in the classroom to teach engineers and computer scientists how to think more deeply about the issues. I mean, look, if unless you give them a course in that expanded understanding of safety, like, what is it? Why would you expect them to solve the problem? They're busy learning about torque and, you know, databases and operating systems. And I don't know, gears and pulleys, like those were the courses I took. Physics and electromagnetics. Like there's no ethics in that, right? Like that's how it's treated. So we need to show them where those connection points are between ethics and the concepts and theories that they're learning about on the technical side. And that's what we're doing in the lab. So at CRAiEDL, that's what we spend a lot of time doing is try to think about how to make those connections and build the tools out and test them and see if it helps or not. And, you know, we're finding that it does no surprise you teach people about ethics and then connect it to their day to day work and give them some tools to get the job done. Why wouldn't it help?
Elizabeth Dubois: [00:32:27] Exactly.
Jason Millar: [00:32:28] The job is like really trying to get the word out and get people thinking about this as like just a part of their day to day responsibility as engineers and computer scientists.
Elizabeth Dubois: [00:32:37] That's great. That's really great. And we have our show notes where there's a bunch of resources that we link to. We also annotate our transcripts in English and French, and so we'll add links to some of these kinds of things, definitely to CRAiEDL and other resources that are available.
Jason Millar: [00:32:52] Magnifique.
Elizabeth Dubois: [00:32:53] We are at time. So I want to end with the final question that I've been posing to every guest this season, and that is, what do you think policy makers are missing right now? Like, what's the one thing that needs a little more attention from the policymakers in your view?
Jason Millar: [00:33:10] Oh, so many things I'm going to pick. I'm going to try to pick one that's most related. If you think the incentive structures are in place for engineers to get this right, just because they're building the technology that does these things like that creates these risks, you're wrong. The history of engineering, like we wouldn't have seatbelts in cars probably if not for regulation. And so getting it right without regulation is it's kind of a fool's game. Like it's just not going to happen.
Elizabeth Dubois: [00:33:40] Amazing. Thank you. This has been a really great conversation. I appreciate your time.
Jason Millar: [00:33:44] Thank you. It's been fun.
Elizabeth Dubois: [00:33:45] All right. Thanks for listening. That was our episode on ethical engineering. As always, we've got a bunch of links to different resources in the show notes, and you can head over to polcommtech.com for annotated transcripts available in English and French that link to a bunch more resources and further reading. This season of Wonks and War Rooms is supported in part by the University of Ottawa's Office of Vice President, Research and Innovation, my University Research Chair in politics, Communication and Technology, the Faculty of Arts and the center for law, Technology and Society. I also want to acknowledge that I'm recording from the traditional and unceded territory of the Algonquin people, and I want to pay respect to the Algonquin people acknowledging their long standing relationship with this unceded territory.
Comments