Can Artificial Intelligence Be Xenophobic? Conflict and Ethical Challenges of Neural Network Development

The article examines the conflict and ethical challenges associated with the “the trouble with bias” of neural networks. Based on neural networks bias conceptions by Kate Crawford, Ezekiel Dixon-Roman and Luciana Parisi, a complex of natural language research for programming, the theory of cultural violence by Johan Galtung, the works by John Rawls and Will Kymlicka on the problem of justice and inequality, the author suggests that the experience of analysis of digital algorithms based on neural network technology has allowed us to form a corpus of ethical colored problems, the central of which is the “the trouble with bias”. “The trouble with bias” is directly related to the phenomenon of “natural language”, that is, a reflection of linguistic practices of the past and present, an average value of the digitized experience of mankind, which is the main resource for learning neural network algorithms. In this context, the article focuses to the stereotypes, power hierarchies, inequalities and discrimination encoded in natural language, which can be described in terms of “classification politics” and cultural violence. The article shows that that the ethical and conflict consequences of “the trouble with bias” can be reduced only if a particular social group receives the status of discriminated. On the basis of the identified substantial characteristics of the current neural network algorithms, the author concluded that there is a high danger of indoctrination of artificial intelligence algorithms with elements of xenophobia, exclusivity and “call-out culture”

Artificial neural network · Computer Science · Ethics and Social Impacts of AI · Neuroethics, Human Enhancement, Biomedical Innovations · Psychology · Artificial Intelligence