One of the growing challenges of artificial intelligence (AI) is the need to create scalable solutions without simultaneously over-scaling reputational, regulatory, and legal risks. This has created a need for ethical AI programs to mitigate these risks. Liability examples include a lawsuit by Los Angeles against IBM for misappropriating data, a regulatory investigation of Optum for an algorithm recommending healthcare providers prioritize white patients over sicker black patients, and regulatory scrutiny of Goldman Sachs for an AI algorithm accused of gender discrimination.
Questions of “data ethics” and “AI ethics” that were once largely reserved for nonprofit organizations and academics are now being addressed by giants like Microsoft, Facebook, Twitter, and Google. These companies are assembling fast-growing teams to tackle ethical dilemmas arising from the widespread collection and use of massive data troves, especially for training machine learning models.
They recognize that failing to operationalize data and AI ethics threatens their bottom line, leads to wasted resources, inefficiencies in product development and deployment, and potential inability to use data effectively. And, there are examples of these Tech giants scrapping AI programs based on ethical considerations (which may have once been almost unthinkable)!
For instance, Amazon scrapped an AI hiring tool due to systematic discrimination against women, and Sidewalk Labs faced backlash and financial loss over its smart city project in Toronto.
Despite the high stakes, most companies still approach AI ethics through ad-hoc discussions, lacking clear protocols. This can result in overlooked risks, last-minute scrambles, or hoping problems will resolve themselves – and to both increases in liability and reputational risk and decreases in stakeholder confidence and employee buy-in!
Here are five information governance best practices that can help companies seeking to minimize the ethical impact of their AI programs and operations:
Leverage Existing Infrastructure: Use the power and authority of existing infrastructure, such as an information governance steering committee, to integrate ethical concerns into data and AI strategies. If such a body does not exist, create one with ethics-adjacent personnel and external experts.
Develop Tailored Ethical Information Management Frameworks: Create records and data management frameworks that articulate ethical standards, identify stakeholders, and recommend governance structures that stress the need to maintain records based on applicable privacy and security laws and standards and to defensibly dispose of unneeded records that have reached their legal retention period (e.g., expired employee reviews). Tailor these frameworks to your specific industry or division/department, such as finance or healthcare, to address unique ethical concerns and ensure maximum applicability.
Learn from Healthcare Ethics and Comprehensive Privacy Laws: Apply concepts from healthcare/HIPAA and comprehensive privacy laws like the GDPR and CCPA, like informed consent and privacy, to data ethics. Ensure that users of your AI systems are informed clearly and early about how their data is used, fostering trust and aligning with ethical standards and to promote information governance by design principles that prioritize the application of best practices for records management into the design of enterprise content management (ECM) and other similar systems.
Provide Granular Guidance for Product Managers: Equip product managers with the knowledge tools needed to balance ethical considerations and business needs. For example, develop tools to evaluate the importance of explainability in AI outputs, tailored to specific product contexts. And ensure that training in ethical AI is relevant to their core job requirements. Also, regularly test knowledge and solicit feedback on the efficacy of guidance and training that you provide.
Build Organizational Awareness and Incentives: Educate and upskill employees across departments on AI ethics principles. Financially incentivize ethical behavior to ensure a comprehensive understanding of the company’s ethical framework and encourage ethical decision-making.
By adopting these best practices, companies can mitigate ethical risks and legal pitfalls and increase the chances for AI technologies to be developed and deployed responsibly.
Comments