“Today, data systems and algorithms can be deployed at unprecedented scale and speed. Unintended consequences will affect people with that same scale and speed” (https://medium.com/ideo-stories/data-ethics-and-ai-276723a1a2fc)
Ethical guidelines for the use of artificial intelligence (AI) algorithms are key for mitigating these risks and this is why:
- AI algorithms are not evil nor are their users. The risks are mostly in the unintended consequences of using AI algorithms. It is one thing if AI is being used for planning purposes but it is another if used to make deterministic decisions about individuals. Ethical guidelines can specify the use and the override mechanisms available to those impacted. This will also protect from knee-jerk reactions to use the AI algorithms for something they were not intended to be used for.
- Predictions are only as good as the data we use. Data in the vast majority of industries are biased because it represents historical human decisions. AI predictions will have these biases baked-in. But “aggregated” across a large number of decision makers and over time may be less baised than any one person. The question is, therefore, whether the AI algorithm is more or less biased than the “business as usual” approach prior to the use of AI. In other words, AI algorithms have the potential of reducing racial and ethnic disproportionately. Ethical guidelines can help ensure the right trade offs are considered and documented.
The usefulness of ethical guidelines depends on making them public, accessible (easy to find, easy to understand and jargon-free) and concrete. Ethical guidelines for using AI is what we need to demand and what we should expect from government and firms asking for our business.