On September 24-25, 2021 Aspen Institute Italia, TIM and Intesa Sanpaolo organized the international conference “Ethics and Artificial Intelligence”, under the High Patronage of the President of the Republic and with the cooperation of Aspen Institute Germany, Institut Aspen France and the Academy of Sciences of Bologna Institute.
The conference, held in the city of Venice, which took its direction from an Honorary Committee, was attended by participants both in-person and online from numerous countries. A Steering Committee of twenty-four experts launched a reflection in 2020 whose results are collected in what they have entitled The Way To Venice Report. The title is meant to suggest an itinerary, this conference being one step along the way, and to supply participants with material for debate. The Steering Committee’s analyses were collated by a group of researchers and later supplemented with the assessments of a Business Users Advisory Board made up of representatives of fifteen companies and business associations that use artificial intelligence (AI) applications. The Report is thus anchored, on the one hand, to the main challenges lying at the point where ethics and AI applications intersect and, on the other, to what industry considers the most important implications for its activities.
The conference sessions were divided into four key themes, as is the Report:
- The first session, “Artificial intelligence and ethical challenges for business and the economy”, examined various aspects that included AI bias and impact on workers. While it is impossible to accept the even remote possibility of discrimination being built into an AI model, AI models are nevertheless vulnerable to biased decisions that work for or against a given individual. Especially if these models are not accurately designed and “trained” on non-homogeneous data. Undeniable is the need to supplement research and industrial policies with a template for an ethical, reliable and efficient approach to the use of AI. This because ethics is the element most capable of fostering industrial transformation without running into social conflict.
- The second session, “Regulating Artificial Intelligence, between freedom and human dignity”, highlighted the problematic balance between the ethical development of innovative technological applications and business development and regulation in a competitive global context marked by differing values. Ethics and rules should not be perceived as limiting AI technology development, but rather as an opportunity. An economic system underpinned by efficient legal constraints (such as intellectual property rights and patents) develops more fully and better than one without them. The concerns raised, essential to the functioning of democratic societies, ranged from incentives to social acceptance and policy. Nevertheless, regulation cannot be immediate or overly stringent, which would place the majority of European start-ups that work with AI at risk of failure.
- The third session, “Technical and moral limitations of Artificial Intelligence in the business environment”, stressed the importance of designing AI models that follow the principle of “ethics by design” – i.e. with a specific focus on ethical practices. This must be the foundation of all the principles, guidelines and ethical codes, business governance, instruments and, finally, persons involved. It is precisely the trust of the latter of these – suppliers, customers, employees and citizens – that must underpin a company’s AI strategy at its deepest level. An example of this is providing the maximum transparency on how data are gathered and used. It was in that same vein of greater transparency that the topic of the need for more white-box testing emerged, as opposed to wider spread and higher performing black-box testing models, which display a complex interpretability dynamic. Ethical practices can resolve many problems in advance of the sale of a product, while regulation on the other hand can provide back-end control. The accent was also placed on the issue of liability when an AI-based system is found to be responsible for criminal activity.
- The fourth and final session, “Education and data for digital citizenship” focused on the separation of science and the humanities in the national educational system as outdated, and the need for a new unified model capable of elucidating their inherent synergies. It is equally important to encourage the development of soft skills and critical thinking precisely with a view to developing the ability to use technology so as not to be used by it.