Building trust in workplace AI

This is an abc podcast.

Has your workplace been transformed by software programs? The work once performed by many people, now replaced by artificial intelligence. AI can be a marvellous co worker who making your life easier. Alternatively, you might feel that an algorithm is now the boss of your working life. Workplace analyst Professor David de Kremer says that organisations too often focus on the technical capabilities of AI rather than understanding how their workers relate to it. He concludes that to get the best out of AI, organizations need to build trust and a positive workplace culture. Paul I'm Paul Barclay in this Big Ideas, Human Workers and Artificial Intelligence. Will we ever trust AI like we do our human co workers? David dicremer gave the keynote address at the Trust Ethics and Governance Alliance Summit at the University of Queensland. He's a professor in Management and organisations at the National University of Singapore and the director of the Centre on AI Technology for Humankind.

This is an important shift that we're seeing today. We're actually living in very exciting times, not only because of everything that's happening at the economic or political level, but also what's happening with our organizations. I always jokingly say that in the past organizations used to exist of people. Interactions among people were the most important. Today, in the minds of many executives, because of the label big data, we've moved to a kind of expression where we would say organizations today consist of data. And of course that is true to some extent, we have so much data available, but on the other hand, there's also a few risks that come along with this when we think like this, because when we start focusing on humans and interactions, we know we have a humane approach. When we start looking at organizations and interactions among people in terms of data, something else may happen, but it is reality today. And just to give you a few statistics of what is happening and where we are today, so PwC and this is one of the more influential references being used, and this was just before the pandemic, because most research after the pandemic is not there yet or it's very much in chaos to say so. But what we saw is that AI will be used at the larger scale and by 2030 it's predicted it will boost our global economy by almost US$16 trillion. I would like to add one thing already, however, is that we hope and expect that it will be used at a larger scale. But what we see today is still that not Even more than 15% of companies are able to go beyond the pilot stage of AI adoption. So there's still a lot of work to be done but these are expectations in terms of importance of AI and how it will matter to our workforce and organizations. Interviewing CEOs worldwide, PwC revealed that 63% of respondents agreed that actually AI will become more important and have a larger impact than the Internet. The Internet that was launched in 94 and 95 to the general public changed our lives. So we are having very high expectations when it comes down to AI to consider even more important than the Internet. And of course, all these expectations, all these forecasts lead to the conclusion that it's inevitable that our workplace will change. We'll see. AI will become the new co worker and the way we will work will change forever. We're already feeling this when we're talking about hybrid working, going back to the office, yes or no. But it will be completely unrecognizable by 2040. That's the prediction. And why wouldn't we? Because if you look at AI and especially those who are on social media, you see a lot of these videos, announcements on what AI is capable of, usually in a laboratory setting because it's not interacting with people. But it seems to be amazing. How are we going to compete with that kind of technology, that kind of intelligence, because we call it artificial intelligence. As a side note, some people are trying now actually to push back on this and regret that intelligence was used for artificial intelligence. But right now we still use that concept of intelligence. And just as a quick introduction to AI itself, AI is actually general label. So most of us talk about AI as if it's a thing, it's a tool and it's all the same. But it's not. So there are different levels to AI. And I came up with this idea that I called the AI Matroshka, the Russian dolls. And once you take dolls out, it always ends up with a smaller doll and then again a smaller doll. Well, AI to some extent is the same. So we have AI as a general label. The big mama, you could say. Within the artificial intelligence we have machine learning. Supervised unsupervised machine learning and reinforcement learning can also be seen as part of this where basically in the case of supervised and unsupervised learning, we provide training data sets. So this is also the type of AI that is used most often within organizations. Remember, organizations consist of data. So we have this data and we provide it as training. We do it either in supervised ways so we know completely what's happening. No black box or unsupervised. We set the parameters, but then the self learning starts and we don't know what is happening anymore. We don't know exactly which model AI is using or reinforcement learning, basically rewards and punishments. It's like a baby is interacting with the environment. It's much more of a biological agent, but it's all focused on machine learning, learning itself from a certain basis. If we go a little bit deeper again, then we see neural networks where we basically try to mimic the human brain. And I'm saying, and I'm emphasizing basically because actually we only know about 15% of how our brain works. So I always tell computer scientists it's a little bit of an exaggeration to say it mimics the human brain if we don't even know how the human brain really works. But in a way, and for people who know statistics like multiple regressions, I compare it to a series of multiple regressions at different layers where we have data, we try to predict and we go and we rerun these regressions at deeper and deeper levels to arrive at a more complex reality to say so. And then finally we have deep learning, which we don't use this at all in organizations or with our customers because it's too unpredictable. We don't know what is happening. And also, deep learning is pretty much debated today. I already refer to the fact that the ones who came up with the label artificial intelligence regretted that they used the concept of intelligence. Well, there's a similar discussion going on with deep learning where people say, is this really a concept that's only applicable to machine to say so humans also engage in deep learning. So just to summarize very briefly, deep learning is still very chaotic in terms of the scientific area, but also what we know about it. But just so you know, AI is a general label, okay? If we see AI is expected to have such an impact, has the accuracy, has the potential and the ability to deliver on big data sets, seeing what's there, seeing what the kind of relationships are way beyond our human capabilities. Of course, it's something that is very useful to organizations and as we all know, corporations, they like to do their cost benefit analysis. The reason why we see it's artificial intelligence. Oh, it could replace to some extent human intelligence. Just like a steam machine or a car replaced our physical body, maybe artificial intelligence is there to replace our human mind. Of course, some of you may wonder now, so what about us then, if our body and mind has been replaced? But it's a very rational approach. This is how it would look like. A cost benefit analysis basically leads to what I call a replication attitude where we see that over time, the more that AI will develop, humans will become more and more obsolete because the idea is there, the intelligence seems more or less the same. AI will be able to take over the job. It will reduce our costs. So at the end of the day, at the top of the pyramid, artificial intelligence will be there and humans will be in service of AI. Some of you will probably say right away now already this is not really human centered. If humans are supposed to be in service of AI, isn't it the other way around? Well, this is how it would look like when we talk about replicating and when we use a cost benefit analysis. So the implicit assumption of such a rational approach are that AI and humans have the same abilities. We both use the term intelligence, but it presents a zero sum frame where basically your loss is the gain of machine. So when machine takes over your job, you lose. And this has led to the idea of autonomous intelligent technologies will be able to be placed in organizations and over time will take over our jobs. And it's not just part of the job, it's just an autonomous technology that is functioning without any human intervention. However, this is a topic that I've been thinking about a lot with some of my collaborators and one of them was Gary Kasparov, the former Grandmaster Chess. Where we were looking at is automated management really an effective one? Just imagine that we are management systems, which basically started with, and a lot of people don't like to refer to him, But Taylor In 1911, scientific principles of Management, where we said the system is holy and humans are unpredictable. And we haven't innovated anything of this because basically our management systems are still ticking boxes, bureaucratic, don't rock the boat. So we use so many metrics, so many KPIs, so basically it's perfect for algorithms to move in. I call actually the management structures we have today the perfect prey for algorithms. And I call it the new MBA Management by algorithm. So autonomous intelligence seem to be able to run our organizations in terms of management. And there's some evidence that people see this happening as well. Again, a survey of Deloitte's showed that Executives worldwide in 2019 expected automation to increase their workforce capacity by 27%. More recent statistics shows that this moves into the direction of algorithmic management. However, just like there's a problem that I see that organizations cannot upscale AI, there's also a problem that most companies don't seem to have a real good business case about what the effects are once we start implementing more and more autonomous technologies. So for example, Deloitte showed again as well that 44% of organizations have no idea how their automation strategies will affect the functioning of their workforce. And that is a serious problem. So I just want to give you one of the more famous examples why it is a problem where. Well, we all know Amazon, and two, three years ago, stories emerged that Amazon was using autonomous algorithms to supervise humans working. When they're packaging, yeah, they basically have to do the work of robots. They're being supervised on the supply lines, they're being supervised by algorithms in terms of their work pace, their, their work structures. So on these assembly lines, of course, it's pretty repetitive work, it's pretty easy, it's a closed system, it's pretty easy to put the algorithm there. But what shocked most people was the algorithm was, under certain circumstances, even allowed to fire the person without any human intervention. So because why? Well, they're looking at the statistics, pace of work, what is happening. And the pace of work goes up based on how you perform. And it's very robotic. I mean, as humans, we have sometimes good days, sometimes bad days, but these good days make up for the bad days because we're more creative or we design things that else would have taken a few days. So as humans, we achieve a balance, a certain equilibrium, whereas a machine thinks very linear and just adds up the regression analysis as already explained. So basically, you create a work environment that feels robotic, and that's exactly what happened at Amazon. So you see this picture of this woman outside, clearly protesting, we're not robots. She's holding a sign saying we're not robots. Also on a shirt. Well, people came out saying, like, I can't work in this kind of environment anymore. I mean, some of the truck drivers of Amazon were known to have all these plastic bottles in their car because robots don't take toilet breaks, do they? So Amazon changed completely the way people had a relationship with their job, the working conditions, but also the mindset and the expectations that come along when we start automating and come up with a certain way of working. And that doesn't seem to be the very nice to humans. So I tried to examine this a little bit more. So in Singapore, I went to tech startups, 74 tech startups to see how humane they think bringing autonomous technologies would be. So this is a very conservative test because they're basically developing the technology. So I asked questions like, development of new technologies will make people less tolerant towards human failure, more tolerant, or it won't make a difference. As expected, even they said that 60% of them said it will make organizations less tolerant towards human failure. Look at the Amazon example. Another question was whether developments of new technologies will make people see humans as less reliable than machines. More reliable or no difference. Again, Here we see 51% of them said it will make people see less reliable than machines, of course, because can we compete with this robot arm in terms of accuracy, in terms of delivery, in terms of pace? We can't. So I also went out to some executives here in Singapore, 239 of them we interviewed and asked them several questions on what will AI autonomous technologies do to your workforce? I asked questions as the more jobs will be automated, the more I fear human employees will be treated like a machine. 59% said yes, the more machine will be used and degree of automation increases, the more I think the pace of work will go up. 76% said yes, the more communication within the organization is automated. And this is my personal favorite because we get so many auto generated emails at our workplace. You know, the ones that don't allow you to respond. The more you get these emails, that's the kind of communication taking place in your organization, the less I feel treated as a human. 77% said yes. So it's pretty clear automation changes the workplace. It changes it for human employees in a way that people feel. There's a lot of negative expectations because you'll be treated like robots. It changes the relationship with the job. They have their commitments. There's basically fear it's not safe for humans. Our data seem to support this as well. And this is where it relates to building a trustworthy workforce. We all know trust matters. Trust is social glue for organizations. It keeps people together. Of course, some critics may say, hey, but earlier on you said organizations don't consist of people anymore, they consist of data. So do we still need that social glue for organizations to work effectively? Well, what I'll show in the remainder of my talk is yes, we do need it. We do need soft skills like trust building and we'll need it more than ever. Just having technology is not going to make us more effective on the long run if we harm the social fabric of our organizations.

Professor David decremer from the National University of Singapore. You're listening to Trusting Intelligent Technologies in the Workplace of the Future, the keynote address from the Trust Ethics and Governance Alliance Summit at the University of Queensland.

So important questions that have come up because AI will have such a big impact on how we will perform, on the organizational culture, on the mindset that we adopt towards work, can we perceive AI as trustworthy. And then an important question is really related to if people don't feel safe and automated technologies are more adopted, are people motivated to trust AI? Because that is important, because they will have to work together with AI to make organizations function effectively. And are they motivated? Well, if you look at the scientific literature, well, we're pretty negative still at this moment because there's something that we call algorithmic aversion. Algorithm aversion. So people are aversive of algorithms. I mean, they look up in awe to technology, usually just like you did with the robot arm and the bowling ball. But we're afraid of it as well at the same time. And the aversion there is that we are afraid that we will lose control. Basically, they will take over, which is especially the case with autonomous technologies. Of course, there's also something which we developed, a mere human heuristic where we found in a series of studies where as soon as it's known that the decision maker has a human element. So just by saying human is making the decision, compared to, for example, a human who is humble, who has so much experience, will make the decision, well, that didn't make a difference. The mere fact that you said human was already enough to evaluate the decision maker more positively. So that shows that we have in our brain a very quick recognition of is it human or non human? That is the splitting point there. So hence we call it mere human heuristic. So based on these people, we expect not to trust AI that quickly. So when we look at this question, are people willing to trust AI and should they trust AI? The way we look at trust in AI today, from the corporate point of view, the governance point of view, when we look at trustworthy items that are being used to assess whether technologies are fitting the bill, we look at trust as a technological thing. We'll say, okay, AI can be trusted and should be trusted. When it's transparent, it's reliable, we can explain how it works. We look at the technological features of artificial intelligence, and if that fits the bill, we can tick the box again, ticking the box exercise. It's trustworthy. It applies to issues like. Or features like, is it open source? Can we explain the inner logic on how data is processed? Which model is being used? Can we alter the settings of the system? The AI system, which means, can we set certain parameters? We have control over it, There's a low rate of failure and there's data privacy, of course. How is it stored? Our data stored? How are data managed? Can we secure that? There's Privacy, obviously, these things do matter, but they're not sufficient in order to say, I really trust technology and I want to trust technology, hence I'll work with it. One of my examples is here in Singapore during the pandemic, we had a token and an app to download, which is called Trace together. So the government knows where you've been in in case you test positive. And it was said, we will only use this data to identify whether you have been tested positive and who was close to you. Then. Then it turned out that they were. That they were also using these data to solve crimes. So there was a lot of fuzz around this. People didn't like it. And when the government came out, they communicated in the following way. No, no, you can trust it. You, you can trust the technology. We made it open source, so you can check it. This is where we wrote several opinion paces saying, well, it's not about whether you can trust the technological feature. What people were complaining about is the way the technology was being used. They didn't trust. So basically it's the person, the institute, the government behind the technology that they were referring to. So the government was talking about, oh, we understand trust was violated. Here's the solution, which is a technological one, but basically the audience, its citizens, they were complaining about the trust of the one who was using the technology and its integrity. So that already showed there's much more to trust in AI than only the technological features. But it's a trend that I see also when we talk about digital transformations. Of course, digital transformations are on everyone's mind. And if it's not today, then you've missed the pandemic, probably because that was the biggest push for organizations to engage in digital transformation. And what I see there in my work as an advisor is companies talk about digital transformation. They only talk about the technology. They forget about the human. But we're talking about a different way of working. But it will be humans who have to adopt that new way of working. So if humans don't want to work with the technology, it's not going to happen. Your digital transformation. But companies are blind to this aspect. And I call this Digital transformations are digitally obsessed. For example, Bosch, where I also work as an advisor, their CEO says, we're obsessed with AI, but unfortunately, what happens is to the exclusion of the human. And I like this cartoon because it illustrates that point. Are you ready to meet about digital transformation? We can't see your slides. Do you need to download the video plugin? We're restricted to this system. Okay. I tried to email, slide, but bounce back. Our File limit is 1 MB, so it goes on and on and on. And you've had these conversations probably as well. Technology limits us in working. So at the end the person says, I'm not sure how to present to you. Can you fax it to us? Because digital transformation excluded humans. The humans and technology have to work together. They and humans have to trust the technology. So trusting autonomous technology, intelligent technologies, goes beyond trust in the technological feature. It goes beyond this idea here that every aspect of learning or any other feature of intelligence can in principally be so precisely described that a machine can be made to stimulate it. So basically, we can make an algorithm create apps for every little skill we need. And this is what we see that tech startups are doing. They're trying to earn money with new technologies, but they try to apply it to any little skill that we find in life. It would be a nice thing because the more we feed the data, the better the algorithm will work. So we can solve everything. But like I said, it's much more than a tech, especially when it comes down to trusting that technology. Trusting also means that people need to be willing to be vulnerable to the fact that they are being paired, have to work together with autonomous intelligent technologies, and that will impact their ways of working, their interests. This is the topic of benevolence and integrity, as already referred to in Singapore. Example, how is the technology being used? So are people, first of all, being willing to be vulnerable? And second, with this technology, is there a certain communication that's still empathic and understanding in a way? Because again, as you saw with the Amazon example, the communication there is very much metric based based on numbers, based on performance, based on efficiency only, which is not the kind of relational communication that people like to see when building trust. Let's try to address these questions. Is this being done? Are we willing, are we willing to be vulnerable to work with technologies? So when are you being willing to be vulnerable? When you know that the other person or agent, in this case the AI agent, cares about me. Well, we have a problem here because I always say AI in a rational, cognitive processing way, can get to know humans. The more data it has, the more observable data, the more it will get to know observations about humans. And on the surface, this is how humans behave. But it will never be a human. You will never understand why it is that you care about certain things. But what is important, it doesn't even care about that. AI doesn't care whether you drop that tomorrow or not. It has no sympathy, empathy in this. It has no benevolence. But that also makes that it doesn't care about your concerns and doesn't understand them. So research shows AI has a problem moving from context to context. If it already has that problem in terms of abstraction, forget it. That AI will understand the different motivations underlying different work situations, which makes that AI doesn't understand cultural sensitivities. And I'll come back to this later. Does AI have its own integrity? Can it actually be ethical? Well, I see a problem there as well. We know that, for example, and this is a case study that we made, AI became a superior chess player to humans. But chess is a game where there's a certain number of rules and there's a number of moves you can make. And of course, the longer in the game, the less moves there are. So in that sense, it's more or less a closed system. But ethics is not the same as playing chess. So on one hand, we have people who think, oh, well, you know, AI became superior in playing chess compared to humans. So AI will likely become superior to humans as well in terms of being ethical. And this is something that I don't agree with, because these people think, okay, because AI is more rational, more consistent, more neutral, it will be more ethical. But does it have its own ethical compass? No. AI has no moral agency. Basically, everything that AI does, AI still needs a human to define a problem, because AI cannot define whether this problem is relevant to you, because it doesn't understand your concerns, nor does it care about it, because it doesn't understand, has no intent, and it cannot make judgment calls. Ethics requires judgment calls at certain points. Certain points, there will be situations we have no information about. And then it matters that you know what the values are and what these values mean to people. AI has no values. It mirrors our values. Again, it learns from our data. So if there's biases in data, AI is also biased. It amplifies it, it enlarges, but it does not understand them as a human being does. So this is important. AI basically doesn't have that benevolence that we expect when people trust each other, nor does it have that integrity when we're building trustworthy relationships. Yes, it has the reliability because of the technological features. Just to give you. I said I was going to come back to the cultural sensitivities. Can AI communicate in ways that foster these relationships with empathy, that it understands us? Well, this is just based on some of the basic research that I wanted to share. For example, AI, it would Be impossible to communicate with people to make sure that they trust you. Because, for example, AI cannot understand analogies if humans talk. Sometimes in the conversation, you'll say, I experienced exactly the same thing. I was in the same situation. Obviously, you weren't in the same exact situation, but all the features fit. You recognize what the other person is trying to say, and you recall that you were in a similar situation. AI cannot do this. They cannot transfer what the sentiment would be in one context to another context, which is the essence of an analogy. Also, AI cannot read in between the lines, which is extremely important when it comes down to understand certain sensitivities across cultures, across teams, across organizations. So these were experiments where they provided the following information. I'll do the same with all of you today. There's three things I'm going to say. A person makes reservations in a restaurant. The person goes to a restaurant, is seated and asks for the menu. The person leaves the restaurant and leaves a big tip. So it's a steak restaurant. Okay, that information I forgot. When we ask the question, did the person eat a steak? Most of you will probably say yes, but AI will say no, because I didn't say anywhere that the person was actually eating. This shows it doesn't understand a context based on the information we provided, but it also doesn't understand what it means to read between the lines. This is where it becomes really important to understand, for example, culture sensitivities. So words are very important in certain cultures. So what I say, that's also what I mean. So this is what we call low context communication cultures like country, where I'm from, Belgium and the Netherlands. Well, what they say, that's what they mean. They're very much in the face. You could say the Dutch. Well, words are very important there. So you don't need to know much about the context. So an AI could probably survive more easily. There's. But then you have, let's say, more Asian cultures like Singapore, China, Indonesia, where we call this high context communication cultures. Words are important, but as you can see in the figure, they're not that important. Other things are important. The tone in the voice, the gesture, the kind of relationship we have. And there it's important to read in between the lines. It's impossible for AI to do this. So this points out the limitations of AI when it comes down to building the trust that is required to work with AI. Because don't be mistaken, even if we talk about autonomous technology, the business model of tomorrow will be a collaborative one where AI and humans will Co create and with my work with Garry Kasparov, relying on his work on chess, basically we came up with what we call triple AI to show that this will be a collaborative model. So we have three types of AI machines. AI, which we call artificial intelligence, people's AI, which we call authentic intelligence, being sensitive to concerns, having benevolence, having an admiral compass, which AI doesn't have. And that will lead to the third type of AI, augmented intelligence, where basically we'll end up with a different figure than the one I started with when we talked about cost reduction, cutting costs. Now the human is on top of the pyramid and AI is in service of it. That's what we call augmentation. So the human is in the loop and we elevate the abilities of humans because at the end of the day, the end user is still human. Because think about it, if we would say trusting AI is an issue of technology, then we indeed put the technology on top of the pyramid because it's on us to understand how a machine works. And once we see on our lists, oh, the machine works as it is supposed to do, we're supposed to trust it. But that's not how it works in reality. It's this. Yes, we need to know about reliability, about intelligent technologies and tech features, but we need something more. We need to be treated like humans still. And that part of trust where the relational aspect, the integrity, the caring aspect comes about, AI is not able to deliver. How do we approach this in organizations? In my book Leadership by Algorithm, I looked at this more holistic approach to trust in the AI era, where of course, we need reliability, like I said, trust in the tech feature. But we need that human element in building trust so that people are willing to work together with the technology. And that's where leadership will play an important role. Leaders will need to make clear today in organizations what is the purpose of using AI and what is the meaning to it, to the organization, but especially to the work context employees are working in. So that's a big responsibility of leaders. And we've done several studies on this. And just to give you two examples, in the first line of research, we showed that when AI is being used in human resource management, in terms of recruiting and autonomous technologies used to make the first shift already in terms of the CVs, do they fit the criteria or when they evaluate your performance in the first stage, what is extremely important is that leaders at that time do show inclusive behavior by listening to some of the concerns employees have, by showing the right kind of values by making sure that that kind of relation is there. Because we found out even though AI was acting autonomously in evaluating your performance, so the supervisor didn't have a direct effect on it. The fact that the supervisor was perceived as inclusive did increase the trust in AI itself. A second line of research where we found that if people trust their organization, they're more likely to trust the AI that that organization adopted. And this was beyond trust in the technological features. So we controlled for this and the reason why was they trusted the organization because their leaders, their supervisors treated them in fair ways, in inclusive ways, in value driven ways, the relational aspects. So this is why I said earlier, it's also the organization, the developer, the people behind the AI that will determine whether you trust that AI. So in that sense, if big tech companies have a bad reputation, it will harm the trust in AI as well. Leadership is important in organizations. If we want to increase the trust in the AI, our people will have to use when engaging in digital transformations. It's not enough to just tick boxes and say, we have all the technologies we need and all the technologies fit perfectly the build. When it comes down to the tech features, you need to do more than this. I always say AI implementation is more than simply saying we have data and what do these data tell us? Extract value from it. And this is a common problem that I see with boards and top executives in organizations. They're afraid of technology, they're not tech savvy enough. So what they do is, oh, computer scientists, tech business services in the company, they'll deal with it, they'll tell us what the data tell us and then we know what to do. Well, if that's the case that you think, then executives are overpaid and data scientists are underpaid. Because you're basically asking data scientists, what do my data tell me? And if once I know, I'll act on it. No, you're being paid for the strategy, not data scientists, because they're not trained as business people. So that means as a leader, you still need to do your job in terms of developing the strategy division, what kind of questions does your organization wants to answer with those data? That's your job still, because that gives the purpose of the organization. So then data scientists know which kind of data do we need to look for? Do you have the right kind of data? If not, then we need to look for new data that can answer your questions. So purpose driven leadership, again is important. We need to integrate our perspectives, we need to have diversity, we need to have an inclusive work Environment to make AI work. Basically, purpose driven leadership and inclusive leadership will become the leadership styles of the future because it will have such a big impact on how we work with technologies and, and to make it meaningful and inclusive in the work environment. And finally, one thing that I've learned when working with aviation companies like Boeing is when you work with data and technologies, you need to have flat communication. When a plane comes in, it has a lot of data. These models are being checked. But the top level knows almost right away what is happening in use of technologies in our organizations. However, because like I said, executives, we're not tech savvy, we leave it up to that level, we don't get that information. Once you engage in digital transformation, communication needs to be flat, more relational. Hence again, pushing that important aspect of human skills. When building trust with AI, it comes down to us humans being clear about why we want to use AI, what it says about the kind of identity and values that we as an organization want to achieve, or we as humanity want to achieve and make sure that AI supports that. That's the augmentation perspective. So to conclude, trust in AI, in the tech features, it is so much more, it's not sufficient. AI is not able to provide the benevolence and the integrity. So we need our leaders in organizations to do that job. Unfortunately, what I see happening with AI entering organizations, I actually see that leaders are taking less and less responsibilities. Exactly. Because they are not tech savvy and they are afraid basically of technology. They think we leave it up to the people who know it and they forget their most important tasks. It's not only about digital upskilling. People don't only need to get to know the technology. We also need to become better at being human at these soft skills that we know we're very good at. So we need to become better at building trust. We need to be better at creating an inclusive work environment, understanding our values, why they are so important, and create a consensus around it. And finally, don't ever forget humans define the problem, set the parameters where AI works within. That means that the one who's behind AI has a bad reputation. It's pretty likely that people, once they know the person, the organization behind the AI has a bad reputation. AI will not be trusted. And a final point that I want to make is I just emphasize the importance of soft skills and building trust is a soft skill. There's new research that shows that we are entering a feeling economy, which means that over time the largest part of your salary will be determined by Whether you have the soft skills and you're good at it and less so whether you have the hard skills and are be able to think in terms of rational ways. Because that's what we have machine for.

Trusting intelligent technologies in the workplace of the future with Professor David de Craymer. While David highlights the importance of soft skills, our education system is currently focused on science and technology. The the STEM subjects. An audience member asks can we turn this around?

How do we convince our political leaders? The mantra at the moment is stem, stem, you have to teach stem. And in fact in Australia over the last few years we've seen a redirection of government funding for student places away from the soft skills to the STEM skills. Not just in Australia, it's in other parts in the world too. So things seem to be going in the opposite direction of what you presented. It's true. So it's a little bit like what you're pointing out is same thing that I'm saying about companies think mainly in terms of cost reduction. And although we know, we all talk about augmentation and human centered AI, at the end of the day we all realize it's still cost cutting, it's still looking at the bottom to see how much money do we save of although we talk about augmentation and it's the same thing, human centered AI today is still humans have to adapt so that pyramid is still there. So in the same way what you're saying, I agree that is still happening. But to give you one answer, I think these jobs of the future are not only stem, but they are stem, pathy STEM together with empathy. So it is both just like it's AI, human co creation, it's both hard skills and soft skills. Yes, people need to understand the technology but not to the sacrifice of our soft skills. And I don't think it's happening everywhere. So for example, in our center we're pushing this thing of the soft skills. And I'll give you an example here in Singapore. So at the beginning of the pandemic, every Singaporean received 600 Singapore dollars for digital upskilling, digital upskilling and human upskilling. So digital upskilling was very clear data management in companies. People have to go through these programs where it's how does an algorithm work? What is AI? How will this change? So it's really only about the technology. So STEM as you said. Now what is interesting, a year ago the government contacted our center and said suddenly they started to realize soft skills were important. So now I have all these people between 40 and 50 in our classroom learning soft skills. So I always jokingly said, yeah, well, here in the Singapore education system, which is really tough on kids already, very early on, it's much cheaper that kids are not treated like an algorithm at age 6, but that they can still play outside, fail, hurt themselves when they fall, learn how to build friendships, be creative. Instead of being told, draw this picture the way it is. No, just draw whatever, because this is not there. It's much cheaper for a government. They start to realize that if we don't know the soft skills anymore, who is going to know? So I always give the example. You can put a computer in a sleeping mode, but I've never seen a computer dream. So basically that means computers have no imagination, humans do. And this is the same thing as saying STEM will probably provide us with 95% accuracy and effectiveness, but without the final 5%, which is the creativity in itself, the 95% is useless. AI cannot identify a problem. Maybe if it reaches a certain level of intelligence, it may be able to do it, but it cannot even do reverse causality. So, meaning you can come up with a solution, but there's no problem for it. So can you find a problem for a solution? That's how post it, for example, was made. Well, that's really a human skill. And this is why I find it important to say these are the opportunities with AI, but these are also the limitations with AI and that kind of balanced education. I agree with you that in a lot of countries they're blind at the moment and walking only one direction. But at the end of the day, it will have to become more balanced again. Stampathy, like I said, because AI is not going to deliver it to us. And some people who say we're talking about general intelligence, that will happen and singularity. Well, there's this recent survey done worldwide among computer scientists who would like to see this. And what was predicted is that not earlier than 2060, the most optimistic computer scientists think there will be general level AI, which means same level as human intelligence, some even say will never be the case. Because as long as we can put awareness and common sense into a machine, our soft skills will become only more important.

David de Craymer, professor in Management and Organisations at the National University of Singapore and the director of the Centre on AI Technology for Humankind. One of the downsides of digital technology is the rise of cybercrime. Those phone calls warning you that your home computer has a security problem, or a text message about a parcel, you, you can't Recall having ordered at the World Science Festival, three cyber security experts discussed what the scams are and how to combat them.

I think the most common one that people will see is phishing attacks. You definitely will get that email where you'll be tempted to click on the link or even give some of your personal information away. And usually it's all in the, the bad guys attempting to be able to get money off you in some way. And I think one of the really interesting things that we're starting to realize is that there's this whole ecosystem to cybercrime that's behind this phishing email. There's groups of people who will actually write the software, then there's a whole group of people who actually generate the phishing emails. And then there's even a whole group of people that'll do the money, money laundering and what happens afterwards when they've actually compromised machines. So it's a pretty large, sophisticated ecosystem.

And so I guess, Bethany, you've sort of had a bit of experience with that ecosystem. I'm presuming with the Australian Federal Police. I mean, it's not just some teenage hacker like you see in the movies, sitting in the basement, programming away, you know, trying to cause mayhem. It's quite a sophisticated group of people.

Cyber crime. Crime is definitely a very complex and challenging crime type for us. It requires dedicated skill sets, substantial cooperation with our foreign law enforcement partners, those kinds of things. You can see the types of individuals that are committing cyber crimes can range from organised criminal syndicates to still that general person in the basement sort of thing that's fallen into in with perhaps the wrong crowd online and is going out and testing their abilities.

Connie, is this an issue for companies as well? I mean, I know I get these texts, but presumably these could be work related, right? I could cause problems at work if I start clicking on random links.

Absolutely. You're spot on there. You know, as Ernest said, phishing is still, you know, one of the most attractive ways of infiltrating a system or, or as an individual, infiltrating your personal information. It's still the largest problem globally that we're faced with. And the reason really around that is adversaries are really cognizant of the fact that humans are the weakest link in any cybersecurity system. The way I look at it is people generally will take the path of least resistance. And so if the human is the easiest way in, they'll absolutely take that. And you know, social engineering has a lot to do with it. You know, People are nowadays putting more and more information online, making it easier and easier to be socially engineered. And, you know, and it just makes it so much easier. You know, you talked about smishing where you're getting those via your SMSs, you know, but there's also, you know, vishing as well, the phone calls, you know, I think you'd be well and truly all have had one of these where, you know, Microsoft's calling you at home without you logging any problems, and suddenly they tell you you've got a problem on your computer. How often does that happen? Microsoft just doesn't call people. So I'm really focused on the human factor. It's something that really fascinates me. I think there's been so many different studies, and one that I always quote is an IBM study where they said 95% of all successful cyber attacks are caused by human error and attackers. We make it really easy for them. They don't even have to do the work, they just send you a link and we infect ourselves. So just by clicking on that and we're making it far too easy. And of course you have your mobile applications which can bring malware into your own phone or your computer if you're not using an app store. One of the easiest way, downloading directly from the Internet and having no idea who's behind, behind that software and what you're doing with that. So, yeah, we're seeing a lot of that.

It's a good thing. I'm very careful about that. I choose my password to be my birth date, of course, which no one would know. So I think that's probably pretty safe. Ernest, can you tell me what would have happened, what would happen if I did click on one of those links? What would typically happen if I, you know, if I wanted that video of me doing something strange and I thought, gee, I really want to see that what would happen. I've never been brave enough enough to try it.

So as soon as you click on the link, it'll try to. Well, normally your computer will try to go to a particular website and quite often that website will be malicious. And really what the bad guys want to do is gain access to your computer and then from your computer into any accounts that you might have. So yeah, as soon as you click that link, the bad guys are looking for that access. So they might get, get access to your operating system, but it might be as simple as just even access to your browser and all the information that your browser might have collected along the way. So they want to use that access to your system to either jump to another computer within your organization to go deeper in and define more juicy secrets, so to speak, or they start looking for credit card credentials, credentials or even just usernames and passwords, because people tend to reuse the same username and password from one account to another one. So if they happen to compromise your library account, so to speak, if you're using the same password for your council library as you are using for your bank account, that's what the bad guys are starting to do.

Bethany, do you have individuals coming to the police with this kind of cybercrime happening? Is it an increasing thing and what sort of information do you tell them? How do you do anything about that?

So where the AFP comes into in reporting of cybercrimes essentially is we're really focused on that unauthorised access on critical infrastructure. I think in terms of reporting of cybercrime, there is a significant underrepresentation of the cyber enabled sort of offences. When you're talking about cyberbullying, the phishing, those kinds of things versus the cyber dependent sort of crimes when you're talking about DDoS attacks and that strictly unauthorised access. When individuals are making reports, there is a collective page they navigate to now and that is report cyber. And that's where we tell victims to input their information into that and then it will be allocated out to the relevant state and territory where that victim is located. I think it's really important in the space cyber is at this time is we're really getting on the front foot of that prevention, messaging and education to the wider public. So hopefully we can prevent the offence occurring in the first instance and not have to go down that investigation pathway.

Bethany Fuller from the AFP with Connie McIntosh, head of network Security Assurance Ericsson and Associate Professor Ernest Thew from Griffith University. You can hear the full discussion on cybercrime by clicking the link on the Big Ideas homepage. I'm Paul Barclay, thanks for your company.

Bye for now.

You've been listening to an ABC podcast. Discover more great ABC podcasts, live radio and exclusives on the ABC Listen.

Apparently.