Iowa State University’s Ivy College of Business has launched the Murray G. Bacon Center for AI Ethics in Business, a new initiative aimed at advancing ethical leadership in artificial intelligence. Led by Professor Michael Howard, the center is a reinvention of a previous program solely focused on business ethics.
The three focuses of the center will be research leadership, student education, and industry engagement surrounding responsible AI use. Priorities in the first year include establishing an advisory board, developing an AI ethics micro-credential for students, and recruiting an interdisciplinary research team.
The Business Record recently sat down with Howard to talk about AI ethics and workplace use of AI. The following is an edited Q&A.
How do you counsel companies to be ethical in their use of AI?
In terms of the Bacon Center and the approach that we take, we consider that there’s a boundary between the digital and ethical worlds, where humans fundamentally have to be involved, because at the end of the day, this is human morality. These are decisions that people need to make. And for that reason, we would recommend that companies design in their systems that use AI, human oversight into anything that’s going to cross into this sort of moral or ethical space of decision-making.
How are people using AI in moral decision-making?
Well, there are good applications and maybe unfortunate or problematic applications of AI in some of these cases. There are a few famous problems recently that emerged. For example, in cases where companies are using autonomous decisions in the human resources space, there have been some incidents there. For example, there’s a case that’s still ongoing with the company Workday. They created an automated system that would do screening of job applicants, and it was alleged that that system has biases in it regarding age discrimination, potentially race discrimination, perhaps because AI systems are trained on historical data, and maybe some of those biases are inherent in the data that was used to train them. There are other cases.
There was an Air Canada case a few years back where they had a chatbot system that was used for customer support, and that chatbot provided information that didn’t really align with company policy. One of the passengers who was flying Air Canada was looking for a refund because they had to travel due to bereavement. They had a death in the family, and the chatbot assured them that that would be appropriate and acceptable. But then the company came back later and said, ‘That’s actually not our policy.’ But basically the AI system had given the wrong information. And so there are increasing examples of these things happening, and companies really need to ensure that they’re not outsourcing moral decision-making in these situations.
Do you think it’s an ethical practice to tie AI use to performance reviews when you’re trying to encourage your workforce to be more efficient?
AI is a great engine for improving efficiency. It can automate certain repetitive business practices that human beings probably really don’t want to do anyway. But if it’s natural for a company to expect or assume that they’ll gain cost savings or efficiency advantages from implementing these things, at the same time they need to ensure that they’ve trained their employees in a way that they have the necessary AI literacy or AI fluency, that they can effectively apply those tools. If you make assumptions about the savings before putting in the work to make people successful, I think that could potentially be an ethical problem.
So you advocate training.
Yes, training and system design in a way that puts people in their proper role in running or overseeing or integrating with an AI system.
What do you mean by putting people in their proper role?
Making the ethical decisions so that we’re not outsourcing that. Human beings have to be this human in the loop that we often hear about. That’s something that I don’t think will ever change. We need to have people in decision-making positions to have the final say when anything involves ethical decision making, and if it’s simply a business process that can be run more efficiently at a corporate level, that firm should provide the training to enable their employees to successfully use those tools and then not hold them accountable in advance before they’re positioned to be successful in that transition.
Do you think that companies should be crafting AI policies to govern the use of it?
Yes, definitely. I also kind of follow the information that’s coming out of Silicon Valley and the high tech fields and things of that nature. There was a Stanford AI Index report that came out last year, and they said there was a 56% increase in incidents regarding AI and that was things that include data breaches to automated algorithm failures that are causing ethical problems, and only 2% of firms have some sort of a comprehensive, responsible AI governance system in place as of last year. However, 60% of companies surveyed said that they think AI is going to be very important in boosting productivity and the ROI of their firm. So there’s a very strong incentive to implement AI but most companies at the moment are not prepared to do that in a way where they have the safeguards in place to ensure that the technology is governed correctly.
How would you advise a company to go about crafting an AI policy?
We’re beginning to learn from companies that are doing well. Take, for example, the issue of data security and privacy. Many companies are experiencing the tendency of their employees to want to work with the tools or experiment with AI to try to do things better. So there’s a lot of enthusiasm around some of their employees for applying AI. But they need to be careful, from a corporate level, not to allow data breaches or or things that are outside of their control. So the very best companies right now are establishing what they would call a sandbox where they get the individual employees who are interested in pursuing AI innovation, and allow them to do so within a system inside the company where the data is secure and then it’s also closely monitored so management can understand that there’s no potential risk that data could be released or that the AI could be misused. But at the same time, they can see the results and the benefits from that innovation that those employees in that sandbox are gaining.
Some people are concerned about how much water is used at data centers. Is that a part of the ethical consideration of how much AI should be used?
That’s definitely a concern. So there’s a wide array of ethical challenges presented by AI. And this is very true in business ethics in general. Whenever you’ve got manufacturing or other large scale operations, there’s environmental impact, there’s the socioeconomic impact for the communities where these companies operate, and then supply chain ethical considerations regarding that as well, in a general sense. But yes, I would say water usage is more of an environmental ethical issue that is connected to AI and increased use of AI. So that’s a challenge that also needs to be solved if we are going to do as the technologists in Silicon Valley predict, ramp up AI capabilities and usage of energy into the foreseeable future.
Do you think individuals should be cognizant of this and use it less?
I think it’s a good practice to focus your AI usage where it provides the most value. And I’m worried that we might become accustomed to going to AI for the simplest questions or the most trivial things, without thinking about the usage and other issues.
How does the center view concerns around AI replacing jobs? Could certain sectors be more affected by AI use?
In the Ivy College of Business and the Bacon Center, that is our highest priority concern from an ethical standpoint, this notion of workforce displacement. The reason is that we have an obligation to prepare our students to be competitive and to thrive in a future economy that integrates AI, and so to be honest, I don’t think anyone you would ask could honestly say they know the answer to how this is going to shake out. For that reason, we are trying to learn as much as we can about the implementation of AI by connecting our students, our faculty and leaders in the Iowa business community to learn what they are facing. Collect this information and then create business cases or white papers that help guide us to the interesting interface between humans and AI. Provide that back to the businesses to help them make decisions and then bring that back into the classroom so that we can teach the next generation of students how to be very effective while using AI.
My own research is focused on technology innovation. I’ve been researching this for decades, and we’ve seen disruptive innovations in the past. I think we’re all familiar with that idea, the personal computer, the internet, cloud computing. I think the smartphone platform had a profound change on the economy and the way that people interact and the way they buy things. So AI is very much along those lines, but potentially more challenging. You were asking, what sectors are more likely to be impacted. That’s one of the big challenges of AI. It has potential to alter the way business practices are performed across industries. There’ll be variants in terms of which ones are hit more than others, but all industries are at risk, or they have the opportunity to improve through the use of AI, so it’s a very challenging situation. I don’t think anyone has the answers, but we need to learn, and we need to adapt to what this technology is capable of doing.
How can businesses or students engage with the AI center now?
The Bacon Center is housed within the Ivy College of Business, and so we’ve just recently pivoted the focus of the center into AI. We will be launching more formally in the fall semester. Businesses could contact the Ivy College of Business, and that will put them in touch with me, and we would be very excited to engage with them, help them navigate this AI landscape, and help our students become connected with the real business world on this profound challenge.