Abstract

During the last decade, the development of Artificial Intelligence techniques for human language processing has opened new frontiers in the analysis of the data generated by users through social media. The possibility of investigating the human behavior with large scale data provides not only new opportunities for research and industry, but also new responsibilities for the AI and Computational Linguistics community as a whole. What is it ethical to infer from social data, where users freely express their feelings and opinions? How should we deal with the biases and stereotypes that are latent in the user generated data? In this paper, we survey the risks and opportunities of analyzing the explicit and implicit affective content in the expressions spontaneously provided by humans in online social fora, with the main aim of highlighting the need to acquire a new awareness of the possible non-ethical uses of automatic human language processing tools and the potential for their ethical uses, such as fighting hate speech.