“ChatGPT, describe the future in 2050.” : Representations of religion and the secular sacred in images of the future created by large language models
Lantta, Jarno (2025-06-17)
“ChatGPT, describe the future in 2050.” : Representations of religion and the secular sacred in images of the future created by large language models
(17.06.2025)
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
avoin
Julkaisun pysyvä osoite on:
https://urn.fi/URN:NBN:fi-fe2025062473602
https://urn.fi/URN:NBN:fi-fe2025062473602
Tiivistelmä
This master’s thesis explores the use of contemporary commercial large language models (LLMs) for creating descriptions of the future. These descriptions are created and then analyzed as individual images of the future, by focusing on how religion and the secular sacred are represented in the different futures.
Three different LLMs are used to create the images of the future: ChatGPT, Gemini and Deepseek. Additionally, three different types of prompts are used to investigate how the LLMs imagine possible, probable and preferable futures and their differences. With 36 images of the future created with each LLM, a total of 108 AI-generated descriptions of different futures forms the data used in the research.
The images of the future have been initially analyzed using qualitative content analysis. Based on both findings from the data as well as relevant literature, six themes have been identified that encompass the representations of religion and the secular sacred. The frequency of how many images of the future include each of the six themes is utilized in the later parts of the analysis, to compare their prevalence to each other and to the findings of previous research.
Selected statistical methods such as Pearson’s chi-square tests and factor analysis are used to confirm and contrast the findings from the qualitative exploration of the images of the future. The research approach therefore belongs to mixed methods research, although the qualitative content analysis remains primary in terms of resources allocated to its execution and presentation.
Some of the central findings of the research include the priority of the secular and the spiritual over the traditionally religious in the images of the future, clear observed differences between what the LLMs found probable or preferable in the future, and some smaller differences between the three LLMs employed. In addition, statistical relationships have been identified where some themes occur together in the same image of the future more often than others.
These findings indicate that the treatment of religion and the secular in the context of the future by the LLMs is biased, which is supported by the conclusions of previous studies on AI bias. The main direction of the bias is towards the secular western societies and worldviews, which is most likely related to the data used in the training of the AI models. The prospective use of LLMs in discussions about the future should take these and other presumed biases into account.
Three different LLMs are used to create the images of the future: ChatGPT, Gemini and Deepseek. Additionally, three different types of prompts are used to investigate how the LLMs imagine possible, probable and preferable futures and their differences. With 36 images of the future created with each LLM, a total of 108 AI-generated descriptions of different futures forms the data used in the research.
The images of the future have been initially analyzed using qualitative content analysis. Based on both findings from the data as well as relevant literature, six themes have been identified that encompass the representations of religion and the secular sacred. The frequency of how many images of the future include each of the six themes is utilized in the later parts of the analysis, to compare their prevalence to each other and to the findings of previous research.
Selected statistical methods such as Pearson’s chi-square tests and factor analysis are used to confirm and contrast the findings from the qualitative exploration of the images of the future. The research approach therefore belongs to mixed methods research, although the qualitative content analysis remains primary in terms of resources allocated to its execution and presentation.
Some of the central findings of the research include the priority of the secular and the spiritual over the traditionally religious in the images of the future, clear observed differences between what the LLMs found probable or preferable in the future, and some smaller differences between the three LLMs employed. In addition, statistical relationships have been identified where some themes occur together in the same image of the future more often than others.
These findings indicate that the treatment of religion and the secular in the context of the future by the LLMs is biased, which is supported by the conclusions of previous studies on AI bias. The main direction of the bias is towards the secular western societies and worldviews, which is most likely related to the data used in the training of the AI models. The prospective use of LLMs in discussions about the future should take these and other presumed biases into account.