Deploy a multidisciplinary strategy with integrated responsible AI

Accountability and oversight must be ongoing as AI models can change over time; indeed, the hype around deep learning, unlike conventional data tools, is based on its flexibility to adjust and change in response to changing data. But this can lead to problems such as model drift, in which a model’s performance, for example in terms of predictive accuracy, deteriorates over time or begins to exhibit flaws and biases the longer it lives in nature. Explainability techniques and human-in-the-loop monitoring systems can not only help data scientists and product owners build better AI models from the start, but can also be used through monitoring systems. post-deployment to ensure that the quality of the models does not decrease over time. time.

“We’re not just focused on training models or making sure our training models aren’t biased; we also focus on all dimensions involved in the machine learning development lifecycle,” says Cukor. “It’s a challenge, but it’s the future of AI,” he says. “Everybody wants to see that level of discipline.”

Priority to responsible AI

There is a clear business consensus that RAI is important and not just an asset. In PwC’s 2022 AI Enterprise Survey, 98% of respondents said they have at least some plans to make AI accountable through measures such as improving AI governance, monitoring and reporting AI model performance, and ensuring that decisions are interpretable and easily explained.

Despite these aspirations, some companies have struggled to implement RAI. The PwC survey revealed that less than half of respondents planned concrete RAI actions. Another survey by the MIT Sloan Management Review and the Boston Consulting Group found that while most companies view RAI as an instrument to mitigate technology-related risks, including risks related to security, bias, to fairness and confidentiality, they acknowledge that they do not prioritize it, with 56% saying it is a top priority, and only 25% have a fully mature program in place. Challenges can arise from organizational complexity and culture, lack of consensus on ethical practices or tools, insufficient employee capacity or training, regulatory uncertainty, and integration with ethical practices. existing risks and data.

For Cukor, RAI is not optional despite these significant operational challenges. “To many, investing in the safeguards and practices that enable responsible innovation at high speed feels like a trade-off. JPMorgan Chase has a duty to its customers to innovate responsibly, which means carefully balancing challenges between issues such as resourceability, robustness, privacy, power, explainability and business impact. Investing in the right controls and risk management practices, early on, at all stages of the data AI lifecycle, will enable the business to accelerate innovation and ultimately to be a competitive advantage for the company, he says.

For RAI initiatives to be successful, RAI must be embedded into the culture of the organization, rather than simply added as a technical check mark. Implementing these cultural changes requires the right skills and mindset. A survey by MIT Sloan Management Review and Boston Consulting Group found that 54% of respondents had difficulty finding RAI’s expertise and talent, with 53% indicating a lack of training or knowledge among current staff members.

Finding talent is easier said than done. RAI is a nascent field and its practitioners have noted the clearly multidisciplinary nature of the work, with contributions coming from sociologists, data scientists, philosophers, designers, policy experts and lawyers, to name a few. a few areas.

“Given this unique context and the newness of our field, it is rare to find people with a trifecta: technical skills in AI/ML, expertise in ethics and expertise in the field of finance,” says Cukor. “This is why RAI in finance must be a multidisciplinary practice with collaboration at its heart. To get the right mix of talent and perspective, you need to hire experts in different fields so they can have tough conversations and surface issues that others might overlook.

This article is provided for informational purposes only and is not intended to provide legal, tax, financial, investment, accounting or regulatory advice. The opinions expressed herein are the personal opinions of the person(s) and do not represent the views of JPMorgan Chase & Co. The accuracy of any statements, linked resources, reported conclusions or quotes is not the responsibility of JPMorgan Chase & Co.

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not authored by the editorial staff of MIT Technology Review.