Moral behaviour in Data Science directly affects the immediate bottom-line as well as how people perceive a business. In fact, there is a debate within the Data Science community on whether it is opportune to introduce philosophers to guide Data Science practitioners in the cloudy field of data ethics.
“If all the philosophers were laid end to end, they’d never reach a conclusion”
- George Bernard Shaw
This quote is true in the field of ethics where philosophers argue about what constitutes ethical behaviour. However, this inertia within philosophy did not stop the attempts to define ethical behaviour when dealing with data. Journals have been issued on this topic, such as the Ethics of Data Science published by The Royal Society.
Despite that Data Ethics may not be well spread however, data scientists and researchers have taken initiative to decide what are the ethical and unethical Data Science projects. Google employees refused to work on military drones and Microsoft’s refused to work on any military-related projects.
Professor Luciano Floridi, from the University of Oxford, identifies and divides Data Ethics into three areas:
Any data ethical policy should start with the ethical data collection. For example, there should be a consent from the audience whose data has been collected. There were breaches concerning this basic ethical consideration, as it happened to Microsoft images scraping for a Facial Recognition System. Realistically, it is difficult to obtain the consent from all the people scraped, however the robots.txt of each website should indicate the parts of the website that can be scraped. If the scraping happens within the limitation of the robots.txt rules, then it’s allowed by the ethical scraping policy.
A different case would be for private data. For private data, the user should actively give consent to use the data captured.
It is unlikely that algorithms are purposely biased, but the data that it’s trained on can be skewed. The latter can create unfortunate accidents, as it happens with one of Google’s image tags. All ethical AI projects must consider the possibilities of bias within data sets. There are tools to assist data scientists to decrease these bias.
Ethical practices are normally set up by governmental bodies such as the “Ethics guidelines for trustworthy AI” released by the European Union and the “Data Ethics Framework” from the United Kingdom Government.
Businesses approach to Data Ethics not only helps to retain skilled workers, but it also protects the company from legal actions, and negative publicity. The most recent example is Facebook and Cambridge Analytica’s scandal. In fact, the activities of Cambridge Analytica not only fatally damaged the company itself, it wiped $100bn off the value of Facebook. Unethical AI and Machine Learning projects will eventually damage the image and profitability of the business.