Dissertation: study seeks ways to overcome ethical challenges in AI software development
Marianna Jantunen's dissertation explores the ethical challenges of AI software production and the opportunities that different stakeholders in AI development can offer in response to these challenges.
"The problem has been the lightning-fast development of AI technology in recent years, compared to which regulations related to the ethical requirements of AI have lagged behind," Jantunen says.
For example, the European Commission's guidelines for regulating AI were only finalised last year, even though AI already has a significant foothold in society.
In practice, AI is now used whenever large amounts of data are processed at once. Jantunen explains that AI is used in many places, such as healthcare, social benefits processing and energy production, as well as in more mundane and familiar places, such as social media algorithms and dating applications.
Jantunen's research explored the attitudes and capabilities of AI developers and researchers regarding the ethical dimensions of AI by asking what concerns they have about different AI scenarios.
"The concerns that emerged in the survey included the risks posed by AI applications in relation to people's privacy and self-determination, as well as the potential bias towards certain groups of people, such as women, that could be transferred to AI," Jantunen says.
Jantunen says that researchers in particular have a deep understanding of the ethical dimensions of AI, but both researchers and developers need effective ways to integrate ethics into software development processes on a practical level.
The thesis offers two methods for integrating ethical dimensions into development work
Jantunen's dissertation offers AI researchers and developers two methods for integrating ethics into application development.
"The first method is the so-called maturity model, which consists of five distinct stages. The model allows a developer of AI applications, for example a company, to assess the maturity of its processes in terms of how well it is prepared for the various ethical challenges of AI," Jantunen explains.
Challenges related to AI can include, for example, discriminating against or favouring a particular group of people. Disadvantages may not only affect users, but also the developer of the application, for example in the form of bad publicity.
Another method is ethical user stories, which is a method based on traditional user stories, whereby the ethical dimensions desired for an application are integrated into the product development process from the beginning, using an ethical framework.
"User stories are written as if they were objectives related to the functionality of the application to be developed, expressing the desired functionality from the user's point of view. These methods are
intended to facilitate the incorporation of ethical goals into the development of AI applications," says Jantunen.
According to Jantunen, the integration of ethics into AI development processes could be improved by, for example, increasing cooperation between researchers and the business community and clarifying the division of responsibilities between developers in AI development projects.
"These measures can give developers a better ability to take ethical dimensions into account in their development work. At best, this will lead to more ethical AI applications in the future," Jantunen sums up.
More information
Marianna Jantunen defends their doctoral dissertation “Integration of Artificial Intelligence Ethics into Software Engineering Processes: Challenges, Concerns and Opportunities” on Friday 28.3.2025 at 12 noon.
Opponent is Xavier Franch (Universitat Politècnica de Catalunya) and custos is assistant professor Rahul Mohanani (University of Jyväskylä). The language of the dissertation is English.
