The Evolution of Ethical AI at IBM: A Conversation with Francesca Rossi

For the past six years, Francesca Rossi has been at the forefront of developing an ethical framework for IBM’s artificial intelligence (AI) technology—yet she prefers not to label it as “ethical AI.” As she speaks from her elegantly appointed home office in Mount Kisco, N.Y., a delicate pink orchid gently bows over her shoulder, underscoring her thoughtful approach. Rossi emphasizes the complexities of the term: “Technology is not ethical or unethical; it’s the whole ecosystem surrounding it.” This ecosystem includes a range of stakeholders—researchers, developers, economists, policymakers, and consumers—all influencing the ethical implications of AI.

Understanding the Ecosystem

At the heart of Rossi’s philosophy is the notion that the goal of AI development should be clear: extract the maximum benefits while minimizing negative repercussions. As companies globally recognize the importance of ethics in technology, they are looking more critically at responsible innovation. A recent report from market research firm IDC reveals that global spending on AI systems is set to rise dramatically—from $85.3 billion in 2021 to over $204 billion by 2025.

Building a Foundation at IBM

When Rossi joined IBM in 2015 as the global leader for AI ethics, she gathered 40 colleagues to initiate a fundamental shift in how AI is regarded and developed within the company. This collaborative effort led to the creation of an internal AI board dedicated to guiding the ethical development and deployment of AI systems. Part of this initiative involved training over 345,000 IBM employees across more than 175 countries in a framework she calls “ethics by design.”

The principles guiding this framework encourage trust and accountability, and draw on partnerships with diverse organizations—from the UN Environment Programme to Citibank, Lufthansa, and L’Oréal. Products like IBM Watson Studio and IBM Cloud Pak for Data exemplify this commitment, offering tools to enhance compliance with ethical AI standards.

Guiding Principles of Trustworthiness

IBM’s strategies are rooted in a long history of technological innovation and corporate responsibility. Since redefining its company values in 2003 to prioritize trust and personal responsibility, the organization has continually refined these to align with societal expectations. The guiding pillars for what IBM describes as “trustworthy AI” include privacy, fairness, explainability, robustness, and transparency.

Rossi acknowledges that accountability is equally vital, noting that decisions made through AI impact not just organizations but society at large. She emphasizes the need for human agency in AI development, especially as these technologies exert significant influence on workforce dynamics.

Accelerating Ethical Initiatives

Since Rossi’s arrival, IBM has accelerated its ethical initiatives, launching programs like Science for Social Good in 2016. This endeavor has included the publication of one of the first papers on detecting and mitigating AI bias and the release of open-source toolkits, including AI Fairness 360 and AI Explainability 360. For instance, the Ad Council has leveraged AI Fairness 360 to identify biases in advertising.

Companies face numerous motivations for investing in ethical AI. From aligning personal values to contributing positively to social justice, equity, and sustainability, organizations understand the reputational risks of deploying technologies that may lead to discrimination. Rossi underscores that no company wants to be embroiled in controversies like the chatbot in France that misguidedly recommended suicide or Uber’s problematic self-driving tests.

The Competitive Edge of Ethical AI

As enterprises recognize the competitive advantage tied to ethical AI, there’s a marked increase in companies instituting ethical guidelines. A Deloitte survey revealed that while 66% of businesses regard AI as crucial for success, only 38% believe their use of it differentiates them from competitors. Furthermore, merely a third report having adopted “leading operational practices” related to AI.

The Broader Move Towards Responsible AI

The conversation around ethical AI extends beyond individual companies. Major corporations like Microsoft and Salesforce are taking steps to create frameworks for responsible AI practices, mandating training for all employees to ensure equitable use of their data and machine learning models.

IBM has actively contributed to this movement through the development of toolkits aligned with its five key ethical AI pillars. These resources—some of which are open-source—help guide companies in making ethical technology choices. Through its Science for Social Good program, IBM also seeks to align AI applications with the 17 United Nations Sustainable Development Goals, emphasizing financial inclusivity and equitable technology deployment.

Navigating International Standards and Regulations

As ethical AI evolves, global standards and regulations are increasingly being established. Recently, the European Commission released a legal framework focusing on the risks tied to different AI systems. This framework not only proposes bans on certain applications but also specifies regulations based on risk levels—challenging the traditionally hands-off approach observed in Silicon Valley.

Moreover, the UN’s UNESCO adopted recommendations aimed at ensuring transparency, data protection, and accountability in AI. Rossi notes that companies operating in Europe will have to navigate these regulations carefully. “Companies need to be prepared and educated about what it means to build or use their technology,” she cautions.

In contrast, the regulatory landscape in the United States remains piecemeal, often varying significantly from state to state. Rossi advocates for companies to adopt a trustworthy AI approach that transcends mere compliance, arguing that “just because regulation allows it, it doesn’t mean that it’s good for your reputation.”

A Future Full of Promise

While the road to ethical AI is fraught with challenges, Rossi maintains an optimistic outlook. Over the last several years, she has witnessed immense progress in aligning AI technologies with ethical considerations. With extensive collaboration and a commitment to continual improvement, the future of ethical AI holds promise for creating technologies that serve society without sacrificing trust.