Time may be right for professionalizing artificial intelligence practices

"It's unlikely that you'd trust a 'citizen architect' to build your home in the same way that you wouldn't visit a 'citizen doctor' when you get sick."
Written by Joe McKendrick, Contributing Writer

With so much riding on the performance and accuracy of artificial intelligence algorithms -- from medical diagnoses to legal advice to financial planning -- there have been calls for the "professionalization" of AI developers, through mechanisms such as certifications and accreditations, all the way up to government mandates. After all, it is argued, healthcare professionals, lawyers and financial advisors all require varying levels of certification, why shouldn't the people creating the AI systems that could replace the advice of these professionals also be verified?

Photo: Joe McKendrick

"For example, you understand that architects, electricians and other construction professionals know how to build a house," says Fernando Lucini, global lead in data science and machine learning engineering at Accenture. "They've had requisite training and understand their roles and responsibilities, safety standards and protocols to follow throughout the construction process. It's unlikely that you'd trust a 'citizen architect' to build your home in the same way that you wouldn't visit a 'citizen doctor' when you get sick." 

A counter-argument may be that there are already too few people well-versed in the ways of AI, and we need everyone we can pull into such efforts. Any attempts at requiring formal certifications or accreditation may stymie such efforts.  

In a recent post and executive brief, Lucini urges organizations movie on their own toward professionalizing the roles and responsibilities of AI practitioners. "Stakeholders -- from practitioners to leaders across the private and public sector -- must come together to distinguish clear roles and responsibilities for AI practitioners; demand the right level of education and training for said practitioners; define processes for developing, deploying and managing AI, and democratize AI literacy across the enterprise. Real value can only be realized when trained AI practitioners are working hand in hand with the business to accomplish their organization's goals, and those interdisciplinary teams are guided by standards, rules and processes." 

While not advocating formally mandated certifications, Lucini says enterprises need to take steps to assure that AI practitioners and their employers follow clear guidelines to assure that AI systems are accurate and ethical. Some leading universities and companies, including Stanford, Coursera and IBM, offer AI certification, as detailed here by Code Spaces.

By formalizing AI as a trade, "with a shared set of norms and principles, companies will be poised to achieve more value from AI," he says. "Yet increasingly, companies are bolstering their core data science teams with 'citizen data scientists,' or people who create models using predictive analytics but whose roles are outside of the data science field, without providing them with necessary guardrails and standards to enable success. Even among trained and credentialed data scientists, there are varying degrees of standards."

Lucini and the Accenture team provide four guidelines for professionalizing AI: 

Distinguish clear roles for practitioners and create additional standards for how to work: "Create multidisciplinary teams of diverse perspectives, skills and approaches who work together to innovate and to deliver AI products or services. Establish ownership and expectations from the start so that AI practitioners understand precisely what they need to deliver and what they are accountable for." 

Demand education and training to create confidence in AI technology, with clear qualifications and standards for practitioners. "Implement regular assessment points throughout employees' careers to test their knowledge and maintain their technical education. Build clear career paths for AI practitioners with pre-requisites such as coursework and training to help build necessary skills and proficiencies."

Take a multidisciplinary approach. "Establish defined processes that formalize the development, deployment and management of AI solutions. Create a standard approach to testing and benchmarking during the creation (or optimization) of AI products and services." 

Democratize data and AI literacy to prepare the workforce for this quickly advancing technology. "Define the minimum level of AI knowledge required from employees - whether or not they work with data and AI as part of their daily tasks. Build a training program that will help build employee knowledge and understanding of data and AI and how it applies to their roles."

Editorial standards