Photograph: Viktor Forgacs/Unsplash
It’s a given that enterprises (if they wish to survive) must shift, grow and embrace new ideas – they must continually innovate. Currently many enterprises are investigating how technologies such as AI can not only speed up the development process for new products and services, but help them define novel solutions totally tailored to the individual.
In industries such as automotive with the advent of autonomous vehicles, the ethical implications of AI are front and centre: if a vehicle must swerve to avoid a person but in doing so hits another person, how can the decision be made as to which life should be saved?
These ethical dilemmas have moved out of the realm of the theoretical, with a recent report from the European Parliament noting “within the last 5 years AI ethics has shifted from an academic concern to a matter for political as well as public debate.” Issues such as racial and gender biases, deepfakes and errors in facial recognition are already transpiring, impacting both individuals and organisations.
Businesses can see the vast potential that AI can offer them. Many enterprises are moving forward and implementing these systems, but often with less-than-adequate forethought for how they will be set up.
As businesses, we not only have a social responsibility to create and continually refine systems that do everything possible to promote equality, but also a duty to customers to explain how decisions made by AIs are reached. The challenge is establishing systems that do not carry with them the biases of the people creating them.
Demystifing AI
Businesses want to innovate and can see the vast potential that AI can offer them. Many enterprises are moving forward and implementing these systems, but often with less-than-adequate forethought for how they will be set up and by whom, how the outputs will be overseen, and how the systems will evolve over time.
The starting point must be establishing ethics guidelines for the business; ethical questions, complex and nuanced as they are, cannot be thought of as having absolute answers. But with guiding principles, for example using the EU’s guidance for trustworthy AI as a basis, which are followed throughout the whole service/product lifecycle, businesses will have the mechanisms in place to catch, and then eliminate, as many prejudices as possible.
AI systems are often seen as an impenetrable black box. At Ensō, NTT Group’s first innovation and co-creation space in Europe, we demystify the process. ‘Explainable AI’ gives insights into the black box of machine learning, and the rationale for its deductions.
Understanding why the AI delivered any given outcome or directive is essential to quantify if the system is trustworthy (indeed, whether it’s trusted by the users is a key success factor). An overall detailed understanding of how the system is set-up, and a clear view of the advocacy of the data being used, are the foundation of trust.
Every person using the AI system to influence their work must trust the data that the system is delivering to them. Trust is based upon a clearly defined set of parameters that the system has been initially set-up to follow, and which are properly maintained and updated throughout the system’s lifecycle. The challenge is to ensure those parameters are ethical and have the lowest level of bias and discrimination possible; and to remember that discrimination is not only along single dimensions (like gender or race), but particularly at intersections (for example women of colour). You can evaluate your parameters by making sure the data used to train the AI model follows both ethical and sustainable data guidelines.
Our clients, particularly those with a more technical background, are excited about using AI – but that enthusiasm does need to be tempered with an understanding that these systems are not like other business applications. For explainable outputs to be gained and to ensure those outputs are ethical and free from bias, a deeper understanding of how these systems operate, and then arrive at their decisions, is at the core of using these systems for ethical innovation.
Innovation without bias
Ethics and diversity are essential components of innovation, especially when using AI. At Ensō, we use workshops to explore these ideas with clients. ‘Ethics by Design’ helps to unravel these challenges, clarifying that although technology itself may be considered neutral, it’s the application of that technology that must be thought about.
What many business leaders struggle with is understanding the level of nuance involved in designing their AI systems, and how their decisions can profoundly impact the outputs they gain. This discussion – about ethics, discrimination and bias, within a framework of technology – can be very new to them. But despite how alien it might feel to discuss philosophical issues, for a machine learning system to be accurate and deliver the tangible results the business wants, these discussions must be had.
In particular, AI and machine learning have many components, all of which impact each other. Unintentional consequences that could be discriminatory and opaque to the viewer are a real risk. We spend some time helping our clients to counteract, for example, unintentional race or socioeconomic biases.
What many business leaders struggle with is understanding the level of nuance involved in designing their AI systems, and how their decisions can profoundly impact the outputs they gain.
Is the data that a machine learning system is based upon sustainable, i.e. is the data up-to-date, correct, accessible, controlled and free from bias as well as discrimination? We often use synthetic data to test the advocacy of an AI system. We then use client data and compare the outputs. These will usually be very different, illustrating the dangers of using machine learning, for instance, without complete datasets. This is often surprising to many of our clients and underlines the importance of using data that has been quantified and analysed before using the machine learning tool.
An ethical tomorrow
Ultimately, for a business to design, implement and then innovate using new tools such as machine learning, the enterprise itself has to change.
People create and analyse the data these systems will use. It’s critical to understand how changes to business culture will have a practical impact on the outputs of the AI system being used.
It is also useful to have an outside view of how a machine learning system is being set up. As we have already mentioned, businesses can often be too close to the datasets they create and can’t see the unintentional bias that could be present.
Ethical standards should be in the DNA of a company to ensure every step taken on the road to innovation is taken thoughtfully… These changes are fundamental to every business that wants to continue to innovate.
We understand that businesses have masses of historical data and continue to collect vast quantities of new data about their clients, customers, competitors and commercial partners. We need to ensure a detailed understanding is present to remove or minimise any bias or discrimination that data may have. That begins with educating a business’s workforce – who are, after all, working with the data the AI tool will eventually use to help them innovate.
Ethical standards should be in the DNA of a company to ensure every step taken on the road to innovation is taken thoughtfully, to ensure high sustainability levels. These changes are fundamental to every business that wants to continue to innovate.
Success by design
Businesses strive to create new technologies, products and services that reshape or even disrupt their markets. Yet businesses also need to understand they must innovate sustainably and ethically.
Equitable design can be product or service focused, but to be truly innovative, companies need to look to the broader society they and their customers are part of. The tools businesses use to achieve that goal are advancing. The insights these new tools are unlocking will be profound. Having a deep understanding of the impact they have on a cultural level pays massive dividends as design and innovation positively impact people’s lives.
