Why navigating the ethics of the digital world is a must-do
ow would you feel if you were rejected from a job in the 21st century because of your gender, specifically because you are a woman? In 2018 Amazon had to discard its artificial intelligence (AI) hiring program because it was biased against women.
How is that possible? The algorithm was trained on CVs submitted for job listings that had occurred 10 years earlier. Of course, this meant the majority of the CVs belonged to men, and therefore the algorithm was taught to favour men over women.
Technology is transforming the world we live in faster than we can keep up. Machine learning (ML) and AI have become trending buzzwords, and most corporates are eager to claim that they are using them in some way. But what is the true intention here? Is it to merely tick a box, or add value? Using technology for the right reasons is important because, if misused, the consequences could be unforgivable (as in the Amazon story).
Digitisation is a wonderful thing; it has opened a realm of opportunities for businesses to progress and become more competitive. Unfortunately, we cannot blame technology for things going wrong; we can only blame the people who implemented it.
We need to be able to ask the right questions. It’s essential that we are critical of the data we use and the information we generate. The extra time and effort used to validate data is less expensive than the issues that can arise from not doing it. Knowing the data’s source, who it belongs to, understanding what purpose it’s used for and what purpose you want to use it for, storing it in the correct format and determining how relevant it is, are some of the answers we need to seek.
This also needs to be an ongoing process in case an anomaly pops up in the data as time goes on. The bottom line is the job is not done after the initial validation; it takes consistent effort and monitoring to ensure data is as accurate as possible. ML is based on learning algorithms, so it can only learn from what it is given. Think of it like raising a child. I have no experience raising children, but I can only assume you would want to teach your children the right things from a young age this is just like designing an ML model.
Designing models ethically also needs to be a priority. The entire process needs to be audited, including the decision-making stage. The intention, goal and limitations should be set upfront, and there should be no deviations from it unless it is to improve or update it. The goal should not be profit maximisation at all costs because this can breach regulations and moral principles.
According to management consulting firm Avanade, “digital ethics sits at the nexus of what is legally required; what can be made possible by digital technology; and what is morally desirable. And therein lies the grey area.” What will allow us to achieve maximum profit is not always necessarily ethical, and this is where environmental, social and governance (ESG) investing also ties in.
Another way we can ensure our models are ethical is to ensure we incorporate unbiased test cases into the process. When training ML models, there is training and testing data. A general rule of thumb when presented with a data set is that 70% is used as training data and 30% is used as testing data.
The training data is used to build the model. The testing data is what the model has not yet seen and thus gives a good indication as to whether the model does well. The testing data needs to be unbiased for the results to reflect the model’s performance fairly.
IT’S ESSENTIAL THAT WE ARE CRITICAL OF THE DATA WE USE AND THE INFORMATION WE GENERATE
Transparency is of utmost importance not only for the data being used but also for the algorithms themselves.
A widespread concern from investors is that algorithms are too “black box” (in other words, the process is not well understood or comprehensible by those who designed it); therefore it is undeniably important to be able to explain and understand algorithms thoroughly. This ultimately creates an element of trust. After all, investors do want to know where their money is going and why.
Navigating the new digital world carefully and consciously is a must, and we need to continuously assess and educate ourselves if we want to remain relevant. It is also important to establish where digitisation will fit in and how it can be used to align with and enhance the organisation’s goals.
This doesn’t happen overnight, as reflected in the following quote by George Westerman, who is involved in the MIT Sloan Initiative on the Digital Economy: “When digital transformation is done right, it’s like a caterpillar turning into a butterfly, but when done wrong, all you have is a really fast caterpillar.”