Metaphors of Artificial Intelligence in Contemporary Philosophy: A Computational Text Analysis of the PhilPapers Database

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F49777513%3A23330%2F24%3A43974135" target="_blank" >RIV/49777513:23330/24:43974135 - isvavai.cz</a>

Výsledek na webu

—

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Metaphors of Artificial Intelligence in Contemporary Philosophy: A Computational Text Analysis of the PhilPapers Database

Popis výsledku v původním jazyce

In this paper, we introduce our analysis of a corpus of contemporary philosophical texts from PhilPapers (https://philpapers.org), focusing particularly on the discourse concerning the perception of “artificial intelligence” (AI) in these texts. Our corpus, downloaded from PhilPapers in June 2024, consists of 67,653 English-language documents. Approximately 70% are articles from academic journals, while the remainder includes books, book chapters, and PhD theses. Each document is accompanied by important bibliographic metadata, such as the year of publication, journal name, and DOI number. The dataset is temporally skewed towards the most recent years, with less than 6% of the texts predating 2000. The most represented journal is Synthese, with nearly 1,000 articles. For our main analysis, we extracted 83,513 sentences containing the phrase “artificial intelligence” or “AI” from 7,370 documents. Given the temporal skewness of the corpus, only 3% of these sentences predate 2000, and 16% predate 2017, the year the Transformer architecture was introduced. We present our computational text analyses of these sentences, focusing on their metaphorical language and sentiment. Utilizing state-of-the-art metaphor detection models, such as MelBERT and FrameBERT, alongside our own BERT-based model, we compare the contextual vector representation of words in sentences to their decontextualized representations. This approach assumes that a word used metaphorically will have a predictably different contextual vector representation. Our study investigates the interplay between metaphorical language and sentiment in the depiction of AI, examining temporal trends and journal-dependent patterns. We explore shifts in sentiment following the introduction of the Transformer architecture in 2017 and trace changes in the depiction of AI from a perfect algorithmic agent to a natural language processing system. Additionally, we analyze differences in AI depiction across various journals, considering national traditions and philosophical subdisciplines. We conclude with a discussion of these trends and their implications for the broader philosophical discourse on AI. Are we observing a significant shift in the portrayal of AI post-2017? How do different journals depict AI, and what might this reveal about the philosophical perspectives on AI across different contexts?Název akce Computational Humanities Research 2024Místo konání Ararhus, DánskoDatum zahájení 03.12.2024Datum ukončení 06.12.2024

Název v anglickém jazyce

Metaphors of Artificial Intelligence in Contemporary Philosophy: A Computational Text Analysis of the PhilPapers Database

Popis výsledku anglicky

In this paper, we introduce our analysis of a corpus of contemporary philosophical texts from PhilPapers (https://philpapers.org), focusing particularly on the discourse concerning the perception of “artificial intelligence” (AI) in these texts. Our corpus, downloaded from PhilPapers in June 2024, consists of 67,653 English-language documents. Approximately 70% are articles from academic journals, while the remainder includes books, book chapters, and PhD theses. Each document is accompanied by important bibliographic metadata, such as the year of publication, journal name, and DOI number. The dataset is temporally skewed towards the most recent years, with less than 6% of the texts predating 2000. The most represented journal is Synthese, with nearly 1,000 articles. For our main analysis, we extracted 83,513 sentences containing the phrase “artificial intelligence” or “AI” from 7,370 documents. Given the temporal skewness of the corpus, only 3% of these sentences predate 2000, and 16% predate 2017, the year the Transformer architecture was introduced. We present our computational text analyses of these sentences, focusing on their metaphorical language and sentiment. Utilizing state-of-the-art metaphor detection models, such as MelBERT and FrameBERT, alongside our own BERT-based model, we compare the contextual vector representation of words in sentences to their decontextualized representations. This approach assumes that a word used metaphorically will have a predictably different contextual vector representation. Our study investigates the interplay between metaphorical language and sentiment in the depiction of AI, examining temporal trends and journal-dependent patterns. We explore shifts in sentiment following the introduction of the Transformer architecture in 2017 and trace changes in the depiction of AI from a perfect algorithmic agent to a natural language processing system. Additionally, we analyze differences in AI depiction across various journals, considering national traditions and philosophical subdisciplines. We conclude with a discussion of these trends and their implications for the broader philosophical discourse on AI. Are we observing a significant shift in the portrayal of AI post-2017? How do different journals depict AI, and what might this reveal about the philosophical perspectives on AI across different contexts?Název akce Computational Humanities Research 2024Místo konání Ararhus, DánskoDatum zahájení 03.12.2024Datum ukončení 06.12.2024

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

60301 - Philosophy, History and Philosophy of science and technology

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů