Ralph Horne Talks With Professor Julian Thomas (Long)

[Start transcript]

[Ralph Horne]

Hi, I'm Ralph Horne. Deputy Pro Vice-Chancellor for Research and Innovation in the College of Design and Social Context. And I've got with me today Professor Julian Thomas, who is the director of the Enabling Capability Platform for Social Change. But also, of course, the inaugural director of the new Centre of Excellence in Automated Decision-Making and Society.

So, hi Julian and welcome.

[Julian Thomas]

Hi Ralph, thanks, great to be here.

[Ralph]

This is really all about this new Centre of Excellence, which is a fantastic thing to have here hosted at RMIT. I'm interested in in the two words at the end of the title, "and society". I suspect, they may be front and centre of what this project is all about and I wonder if you could start by telling us a little bit about how society fits in to this Centre of Excellence?

[Julian]

Yeah, so absolutely. So, this is a Centre which has its own centre of gravity in the social sciences and in the humanities. We are working with colleagues from Computer Science and other technological disciplines. But it really is about the social consequences of these new technologies of automation. What we call, "automated decision-making systems", that we are most concerned. We are concerned with the social consequences of these systems. We are concerned with how they are changing society in all kinds of ways. And we are concerned about how society, in fact, shapes these technologies as well.

So, we can't just understand them as purely technological constructs. They are the kinds of things which you need to take a complex view of. And we think you can only really understand how they are going to play out for us, why are they going to be so significant, if you bring all of those social disciplines and perspectives to bear, as well as the purely scientific ones.

[Ralph]

So, can you give me maybe some examples of how you see this Centre helping us to understand shifts in urban systems arising from kind of automation as described?

[Julian]

So, yes. We see automated decision-making changing the way our cities work and our experience of them. The kinds of problems that we are dealing with on a daily basis. Nobody, for example, decided that our residential apartment buildings should suddenly become hotels. Nobody decided that our quiet residential back streets should all of a sudden become busy roads. But these things have happened, and they've happened because of a specific kind of automation. Because of the development of automated decision-making machines that we call, recommendation systems. Which now tell people where they might stay if they are visiting a city or which way to drive home.

And these have extraordinary effects, because they've been designed to maximise the benefits for their users and for the corporate platforms that develop them. So, public and social interest considerations have been left out of the design of those systems. If we understand them a lot better using work with our colleagues in the technological sciences and our social sciences and humanities to get a better feel for how this is actually playing out on the ground, then we might be able to redesign those systems. So, they actually take into account the fact that there is a public interest in quiet streets, in quiet neighbourhood sometimes. As well as an interest in making sure that people can get home and have places to stay.

[Ralph]

So, um, if I follow up on that. Can I ask you to talk a bit about how the researchers will investigate the kind of uneven consequences of automation that you are talking about? What kinds of research methods do you think across the different disciplines you've got involved will be used?

[Julian]

So, you've got to do a whole lot of different things and this is one of the things that makes the Centre complex. Makes the research program larger than it would be for a regular research project that we might do in just one discipline. Because to understand the problem we've been talking about, as I said, you've got to draw on skills from very specific computational disciplines, data science. And you've also got this problem of how do we understand what is happening? So, for example, our colleagues in digital ethnography who are expert in understanding the uses of technologies and how people adapt them and change them are also going to actually be vital.

Then we also need to draw on our experts in economics, in public policy and law who can look at the regulatory structures-- whether they are formal or informal, embodied in code or embodied in law or in local policies to understand better how they can be modified to achieve the sorts of results we want, which actually balance public and private interests. Because the alternative is just banning these things and we've seen that happening elsewhere. So, a better result is possible. It is possible to apply a public interest set of principles to this, but you need to draw on all of those areas of expertise in order to do that.

[Ralph]

And how will the centre work with, what we might call, end users to ensure that this new knowledge that is created around the consequences of automation gets taken up in public policy and in practice?

[Julian]

Well, our end-users are not end users. They are partners in the research, and they are involved in the design of the research from the very beginning. They are also end-users, we hope. And we hope we will have a lot of end-users. But we are particularly interested in this collaborative relationship.

So, who will be going to work with? We're going to work with technology companies like Google who can help us understand better the way in which these systems are designed, the technologies behind them. We're going to work with legal, regulatory and policy experts outside the universities who are involved in these issues on a day-to-day basis. We're going to work with consumer organisations. Those organisations which are representing the interests of ordinary citizens who are finding themselves often on the sharp end of these kinds of transformations.

But those organisations are often very well-placed to amplify the results of our research, to disseminate and communicate what we're doing. To work through how they might be usefully embodied in policy change or different practices that might be developed by particular kinds of institutions. So, we are really relying on those industry partners from the very beginning, for expertise, for ideas, for access to critical research resources. And, of course, for the engagement and impact which is absolutely necessary for a successful large-scale research program like this.

[Ralph]

And can I ask you a bit about the pace of change? So, we see these apps, these recommenders coming on the stream almost on a daily basis. And we are talking about a complex program of research to try to understand that in anticipation, but also in practice and in circumspect. So, can you talk about the temporal rhythms of this project? How we do the kind of understanding and feed that into decision-making and design in a kind of cyclical manner?

[Julian]

That's right. So, it's a very-- it's a rapidly changing environment and we are talking about a whole mix of technologies. When we talk about automated decision-making-- this is a very important point. We're not talking about a specific technology. This is not just about robots. It is not just about AI. It is not just about the blockchain. It's about all these things. About how they are changing the way automation is going on. They are changing the way our institutions work. They are changing resource allocation and all kinds of important aspects of people's lives across the world.

And that's happening very quickly. The sorts of-- the kind of technology we were just talking about, recommendation systems, they were almost nowhere 20 years ago. It was Amazon which actually developed them dramatically. And then, found that they were rapidly picked up, adopted and adapted using new techniques like machine learning in recent years and distributed right across the world of electronic commerce and daily life. So now, we can't really pass a day without coming into contact with a system like that. They are now in the cloud, they are now extraordinarily cheap. You used to have to employ a horde of highly paid expert computer scientists to develop a bespoke recommendation system if you wanted such a thing in a, for example, university to suggest to students what courses they might be interested in.

Now, it's a simple matter of subscribing to a cloud-based computing service. So, these things are literally everywhere. They are widely dispersed. We can expect that to keep going on with consequences that we don't yet know about. But there are also new technologies emerging, which are more experimental, which are at an earlier stage of development, where the discussion about things like ethics is much less well advanced. So, blockchain is a good example of that. We hear a lot about AI and ethics. We don't hear that much about the blockchain and ethics. But believe me, there are as many problems with those kinds of systems and there are likely to be as there will be with AI.

So, by taking a technology neutral view, the Centre is going to be able to focus on the effects of technological change, rather than the machines themselves, as it were, in order to make a difference where it counts. Which is in how these technologies impact on ordinary people.

[Ralph]

So, stepping back and just thinking about the next 10 years. We've presided in the last couple of decades over kind of widening gaps in inequality and disadvantage, growing disadvantage across the world really in an urban population. I wonder whether automation, instead of being a problem in exacerbating those already existing problems, could actually become an opportunity to address some of those widening gaps and problems? And if so, that this research could inform that. Can you give us a kind of example or two of how that might happen?

[Julian]

Yeah, of course. And, I mean, this is our view. There are multiple scenarios out there for the effects of automation and where it might take us. We are familiar with the story of increasing inequality, declining productivity, lower wages, lower employment. That's a real-- that's a real scenario, it's a real possibility.

We are also seeing extraordinarily-- extraordinary benefits from new automated technologies. So, we are seeing, for example, that, say in the field of natural language processing. The capacity for machines to understand and translate may enable a whole new domain of cultural diversity. It may enable us to keep alive, say, for example, Indigenous languages, which are now facing extinction, because not enough people speak them. Not enough people understand them. And the work and time and resources involved in keeping those languages at the moment is enormously challenging.

So, there are all sorts of ways in which we think the new technologies of automation have great things to offer. But they need to be controlled by citizens. They need to be managed by citizens. They need to be managed by institutions which are operating in the public interest. We find that in health, when we're looking at the development of precision medicine and all kinds of other targeted decision-making systems. We find it in talking about news and media. That is absolutely the case there, where as well as all of the problems that we are seeing with platforms like Facebook and automated advertising, targeted advertising, those kinds of things. We are also seeing a world of options and capabilities for people to make their own media that haven't existed before.

So, we are seeing this in a whole lot of different domains, and we want to make sure that those positive opportunities and capabilities aren't lost to Australians, because we know too little about the technologies. We know too little about their social impacts and we are too afraid of them because of that.

[Ralph]

Thanks Julian. I'm sure we will be talking with Julian on many future occasions about the progress of the new Centre for Excellence in Automated Decision-Making and Society. Thanks.

[End transcript]