Ethical Aspects of Big Data
What can big data be used for? Their use is seemingly unlimited. The terms of machine learning and big data are being used in all possible industries.
Algorithms used in big data processing are able to learn. However, their learning ability is limited to the existing cases they have in their databases. This is advantageous in some situations. The algorithm is thus well adapted to recognize, say, the structure of an emerging cancerous tumor. In situations where more complex evaluation is needed, however, this unilateralism may not produce optimal results. Already at the International ITAPA Congress 2017, futurologist Gerd Leonhard spoke about the threats of handing in decision-making power to technology.
In the hiring process, few companies took advantage of using face recognition technology. HireVue, an American company, has developed a technology that analyzes the words and voices of individual candidates who answer the same questions during a job interview. Assessment is based on such detailed parameters like the use of passive and active words, sentence length, speed of expression, eyebrows, smile, narrowing or widening of the eyes. According to HireVue, the technology that has been already used by 700 companies, allows you to reduce the recruitment process by 90%. The system works with a database of previous interviews with candidates who have succeeded in the process.
Each technology has its limitations. A similar use of artificial intelligence leads to newly recruited employees who are very likely to possess very similar if not the same characteristics as the employees already working for the company. It is seemingly desirable to be able to find people who would adapt to the company culture easily. However, sharing similar political or cultural views doesn§t really guarantee the ideal and conflict-free work environment. On the contrary, the most productive and dynamic teams are usually achieved by different types of people.
Another potential problem might appear if the system will not be able to reliably analyze the types of people he has not encountered before. Suppose the algorithm was trained on a sample of Americans. In a situation when a Japanese person speaking perfect English comes to the interview, there is still a possibility of not passing the interview, taking into an account cultural differences which imply different facial expression, voicing, linguistics and so on. Artificial intelligence takes into consideration data of the successful candidates, but it will not evaluate the success of the candidates in any other way than it was programmed.
This might put us in a situation where artificial intelligence could become a bad master instead of a good servant. In the sophisticated use of AI, we cannot forget about the limits algorithms poses. So far, Kant's categorical imperative and Asimov's three laws of robotics hasn§t been programmed into computers. But as Gerd pointed out at the International ITAPA Congress 2017, it is time for a conversation about the role of ethics in digital technologies.