Keywords: Olympic Legacy; Sentiment Analysis; Discourse Analysis.
Advances in machine learning techniques and the high availability of data and compute power have given rise to a new generation of AI and Natural Language Processing (NLP) approaches, which have achieved unprecedented performance in tasks like question answering and Natural Language Generation (NLG). In fact, NLG engines can create texts so readable that they are capable of deceiving readers into thinking they have been written by a human, effectively passing a hypothetical Turing test. This prompts important questions speaking directly to the core of hermeneutics - the study of meaning and interpretation of texts- which has traditionally relied on a perceived social contract between authors and readers (Henrickson, 2021, p. 4). It has been shown that when text creation is carried out by an NLG engine, the contract holds, with readers still perceiving elements of authorship even in generated texts (Henrickson and Meroño-Peñuela, forthcoming).
However, this perception seems to occurs when we detach AI and NLG engines from their broader societal contexts. These systems are put in place by someone (e.g. a company, an individual, a government), and they are trained on data created and curated by humans. In the general narratives that permeate society, usually rich in hype towards new technologies (Milne, 2020), AI is presented as useful and objective, but this contrasts with the aforementioned acknowledgement of human intervention. Such intervention often focuses on optimisation for profit, with optimisation efforts contributing to AI that is as biased, fallible and subjective as humans. In this paper, we investigate what it means to have NLG ‘authors’, as well as the ability of the hype surrounding NLG and AI to deceive and mislead. We highlight the need to find ways of keeping NLG in line and accountable through regulation, provenance, and dataset documentation.
References
Henrickson, L., 2021. Reading Computer-Generated Texts. Cambridge University Press.
Henrickson, L., Meroño-Peñuela, A., 2021. The Hermeneutics of Computer-Generated Texts. Configurations — Journal of the Society for Literature, Science, and the Arts (SLSA). JHU Press (in press).
Milne, G., 2020. Smoke & Mirrors: How Hype Obscures the Future and How to See Past It. Hachette UK.