Artificial intelligence is an amazing tool to improve efficiency and speed up processes but, when not used properly, can cause more damage than good. With this in mind, how can companies ensure they are using AI ethically and responsibly?
Tip #1: Biases in AI are not always where we might think they are. When most people look for biases in AI, it’s usually focused on unbalanced training data. An example is having more men than women represented in the dataset, or vice versa. This isn’t the only place where an AI bias can happen, though; it can also be when an AI algorithm is used outside the content it was initially created for. For example, a product that is specifically built for farmers wouldn’t have factory workers or business executives represented. This isn’t AI bias, but design bias.
Tip #2: Have a diverse team. Left unchecked, human biases can find their way into datasets, corrupting the decision-making process that the system uses to generate solutions. Having a team that is diverse brings diverse thoughts and opinions, ethnic and cultural backgrounds, ages, etc., and this will not only challenge ideas but can catch biases before they’re implemented into an AI system. Having a diverse team truly makes the difference when striving for a stronger Ethical AI framework, especially when ensuring the avoidance of Simpson paradoxes. (Simpson’s paradox is a phenomenon in probability and statistics in which a trend appears in several groups of data but disappears or reverses when the groups are combined.)
Tip #3: Use bias bounties. One way to detect biases and discrimination in AI is to use bias bounties to catch bad data, avoiding further deviation of the analytics. Bias bounties are implemented to reward users for identifying bias in AI systems. These biases happen as a result of incomplete data or existing bad data that can lead to an AI that is unethical.
Several major companies are using bias bounties as a way to ensure their AI systems are as ethical and robust as possible. It’s an easy and relatively inexpensive way to get more diverse thought processes to look at and review a business’s AI.
Tip #4: Cultivate responsible AI at every level of the company. Having responsible AI at all levels within an organization is a necessity, especially now. Before, organizations could get by with just having their technical R&D or research teams involved, who would then communicate downward to their customers and partners. With the ever-changing landscape of responsible AI use, this is an area where no compromises can be made.
Christophe Bourguignat is CEO and co-founder of Insurance tech provider Zelros.
