Notes
On the Road to Media PsAIchology? Mapping the potentials and challenges of artificial intelligence for research in media psychology
15The recent explosion of large language models (LLMs), such as ChatGPT and GPT4, and tools based on those has led to an increased interest in the development, use, and impact of artificial intelligence (AI) in the media and public discussions as well as within scientific community. As for other digital technologies, the relationships between science and AI are manifold: AI is 1) the result of scientific research, 2) an object of study, and 3) a tool for researchers. The first type of relationship is most prevalent in the field and computer science and, in the case of LLMs, (computational) linguistics. Since studying the use and effects of media and digital technology is a key defining feature of media psychology, this type of relationship is especially important for our field. Notably, however, the use of AI-based tools also has a special potential for media psychology research. While using AI for discovery (e.g., literature research) and data analysis tasks is a promising avenue across disciplines, for media psychology in particular, emerging AI tools have the potential to transform substantial parts of study design and data collection in various lines of research as they can be used for the creation of stimulus material for experimental studies and the coding or classification of media content. Regarding the former, various generative AI models are able to produce high-quality text material and also images based on relatively simple textual prompts. And while the quality of AI-based text-to-audio and text-to-video tools is not yet comparable to those for text generation or text-to-image, we will likely see significant improvements here in the near future. With regard to content classification, there is some initial work suggesting that LLMs can be used for tasks that traditionally required human coders; and with multimodal LLMs this can be extended to combined text and image classification. Besides having the potential of improving the quality of research, the use of AI tools for study design and data collection can also positively contribute to inclusion and diversity in media psychology. AI tools can reduce barriers such as costs and required skills for the production of high-quality stimulus materials or data collection. Independent of the many potentials that AI tools have for media psychology research, however, there are various challenges that we as researchers need to consider and address. Many of those relate to legal and ethical issues, e.g., regarding data privacy and copyright. An associated problem is that of relying on commercial services which has already proven to be risky for data access. Against this background, the use and support of free and open-source (FOSS) projects is a particularly relevant subject. The position paper will discuss the potentials as well as the challenges of artificial intelligence for research in media psychology, paying special attention to the possible impact on inclusion and diversity in our field. Considering the major impact that AI will likely have, the contribution will also discuss whether we must rethink the general types of relationships between AI and media psychology.
16Artificial intelligence / large language models / methods / metascience