This article was originally published on Bloomberg Law, authored by Eric Winston, Executive Vice-President, General Counsel, And Chief Ethics And Compliance Officer, Mphasis.
Using artificial intelligence can result in and perpetuate biases, writes Eric Winston, general counsel and chief ethics and compliance officer at IT services company Mphasis. He provides best practices for businesses, including employing a chief AI ethics officer, to ensure AI does not conflict with established rules of ethical corporate behavior.
As nations across the world slowly reopen their economies after extended lockdowns, businesses will need to hit the ground running to operate in a new abnormal. One of the ways companies can count on meeting the acceleration with safety is by adopting smart tech, especially tools and platforms enabled by artificial intelligence.
However, because these tools and platforms are built on algorithms, there is concern that the use of AI technology might unconsciously result in and perpetuate biases. When it comes to this area, a business’s commitment to ethical operation is a must in a more transparent world where consumers are keenly aware of a company’s track record and business conduct.
What can businesses do to effectively tackle this challenge? How can organizations safely deploy platforms enabled with AI to do more with less while ensuring that they are always doing the right thing?
Ways to Ensure Ethical Adoption of AI
Enterprises can undertake five best practices to ensure the adoption of AI does not go against the established rules of ethical corporate behavior.
First, organizations must have a clear understanding of what practicing ethical AI means to them and communicate this clearly to stakeholders. These communications should convey the core values that define a business, whether it is transparency, customer delight, or people-focus. An ethical application of AI will then mean that none of these values are compromised or watered down irrespective of the corporate function that executes it.
Second, businesses must invest in ensuring a more ethical application of AI by employing a chief AI ethics officer. This clearly defined position would be in charge of overseeing, limiting, and assessing how AI is embedded into an enterprise system. This would help to restrain a company from allowing AI to be used to carry out inappropriate or controversial functions, such as facial recognition.
Take for example the U.S. federal government report that found facial recognition algorithms had a higher false positive error rate at a factor of 10 to 100 times for African American and Asian people relative to white men. As a result, businesses have come under greater scrutiny for their sale of technologies to law enforcement firms that enable facial recognition.
Having an AI ethics officer will also help ensure that companies are on the right side of regulatory compliance with issues around personal data and data privacy. In a recent survey conducted among senior business leaders, NewVantage Partners found that more than half of those polled, 55.7%, indicated data ethics was a top business priority.
Third, companies need to incorporate an AI ethics and quality assurance review as an integral part of the product development and release life cycles, including a focus on various use case scenarios and resulting outcomes. Every new AI-enabled product should be examined from an ethical lens to confirm that it adheres to established protocols around data safety and compliance.
These practices, once embedded in both the development and release teams, can in turn create a seamless, integrated approach toward the introduction of any new AI or machine learning-empowered solution. This will allow for proper vetting for adherence to ethics, GDPR requirements relating to data minimization, adherence to ‘privacy by design’ processes, and pruning for potential biases.
Fourth, enterprises can ensure ethical AI deployment by turning to the customer. A cross-section of experts selected from a company’s advisory council can be leveraged in the testing of newly created AI enabled tools. Their inputs and experience can be funneled back into the product cycle to help the solution become more transparent, fair and impartial, safe, and ethical.
Finally, the fifth way for a business to adopt AI while staying within the confines of regulatory compliance and ethics is to remain transparent about how data is used to build algorithms. Since these algorithms tend to be opaque and complex, an enterprise, while balancing IP interests, may consider going beyond the call of duty to explain and describe to its customers what data is being sourced, and for what purpose.
Making a clearer link between the value data offers to build efficient algorithms and its potential ability to deliver superior customer experience will go a long way in assuaging customer concerns about the ethical deployment of AI to deliver products and services.