The Role of Artificial Intelligence in Shaping Collective Imagination and Reinforcing Cultural and Social Bias

In recent days, social media has been flooded with thousands of AI-generated images inspired by the delicate faces, poetic landscapes, and dreamlike atmospheres that Hayao Miyazaki has made inimitable through decades of artistic research. The latest feature of ChatGPT, allowing users to “Ghibli-fy” any photo, offers the illusion of inhabiting that universe through personalised imagery.
The evolution of AI-powered image generators has transformed the creative landscape, providing tools capable of translating textual prompts into complex visual representations. However, this technology is not without its challenges, which extend beyond copyright and authorship concerns to more profound ethical dilemmas. One pressing issue is that models such as Midjourney, DALL-E, and Stable Diffusion rely primarily on Western-centric datasets, thus reinforcing biases and stereotypes that perpetuate pre-existing cultural and social prejudices.[1] Recent studies and critical analyses have highlighted how these biases are being strategically exploited by right-wing political movements on both a global scale and within the Italian context.
The biases embedded in generative AI not only replicate existing stereotypes but are strategically exploited to reinforce exclusive identities and fuel radicalizing political strategies.
Research has consistently shown that bias is embedded in generative AI models. Studies by VPN Mentor (2023) and a joint investigation by Hugging Face and Leipzig University have demonstrated how prompts such as “happy family” often produce images reminiscent of 1950s Western ideals,[2] while “professor” overwhelmingly generates depictions of white men. These outcomes are unsurprising given that the training datasets privilege Western content, thereby reinforcing a limited and stereotypical worldview.[3]
An analysis by AlgorithmWatch further reveals how AI image generators replicate representational systems that reflect “whiteness” and an overarching Western aesthetic, neglecting cultural diversity and plurality.[4] Moreover, critiques published by MIT Technology Review (2023)[5] and other investigative reports indicate that despite efforts to introduce diversification (such as those seen in the Midjourney 5.1 update) AI models continue to stabilise predefined visual patterns based on user feedback, further entrenching these biases.[6]
Beyond merely generating images, AI functions as a “nostalgia machine,” capable of evoking an idealised, culturally homogeneous past. This mechanism has been strategically leveraged by right-wing political forces to reinforce “collective” and “exclusive” identities. Research conducted by the Institute for Strategic Dialogue (ISD) has shown how European right-wing movements (particularly in Germany) use visual narratives to construct a polarised vision of society.[7] In Italy, too, political discourse increasingly incorporates AI-generated imagery referencing symbols of “traditional Italy,” where rural landscapes, historic monuments (sometimes purely imagined), and community life scenes are employed to counter multiculturalism, which is framed as a threat to national identity. Groups such as Fratelli d’Italia and CasaPound have adopted similar tactics, using AI-generated visuals to reinforce an exclusionary and nostalgic identity, where a mythologised past becomes a tool for both political mobilisation and division.[8]
Another critical concern is AI’s capacity to generate content that profoundly influences the emotional fabric of society. Memes, manipulated videos, and deepfake technology are actively employed to foster polarisation and create neo-tribal emotional communities, where political narratives are experienced as intensely shared identities.[9] Analyses by European think tanks and research centres indicate that the widespread dissemination of deepfakes and altered images contributes to destabilising public debate and fostering individual radicalisation.[10]
In the Italian context, the impact of these technologies has been particularly evident during electoral events since 2018, playing a significant role in the 2022 and 2024 campaigns, where manipulated images were used to amplify divisions and fuel fears, thereby facilitating the adoption of extremist positions.[11]
The growing prevalence of AI-generated images has necessitated the development of specialised methodologies to distinguish them from authentic visuals. Companies such as Prisma Srl (2025) are working on innovative strategies and tools for the identification and verification of AI-generated images, helping to counteract visual disinformation. These methodologies, based on digital forensic analysis, represent a crucial step towards greater transparency and accountability in political and media contexts.[12]
The future challenge is twofold: developing technologies and regulations that ensure fairness and transparency while fostering a critical and informed debate on AI’s role in shaping collective imagination.
The political use of AI-generated imagery also extends into the legal sphere. The European Union has taken a pioneering role in regulatory frameworks, introducing the Artificial Intelligence Act in August 2024 to ensure transparency in training datasets and mitigate algorithmic biases.[13]
In summary, AI-generated images have become a battleground where technological innovation intersects with ideological narratives, with profound implications for political, propaganda, and extremist strategies. The biases embedded in generative systems not only replicate pre-existing stereotypes but are actively exploited to reinforce exclusive identities and fuel radical political agendas, both globally and within Italy. The future challenge is therefore twofold: developing technologies and regulations that ensure fairness and transparency while fostering a critical, informed debate on AI’s role in shaping collective imaginaries.
Academic research and extensive critical analyses conducted by experts across various disciplines highlight the urgency of involving the research community, institutions, and civil society in defining shared guidelines to counter the monopolisation of visual representation by elites and extremist movements. It is essential to diversify training datasets, incorporating sources and data from a wide range of cultural contexts, and to adopt transparent, verifiable algorithms. A concrete proposal involves the establishment of multidisciplinary ethics committees, where experts in social sciences, computer science, and law collaborate to oversee and guide AI model development. Furthermore, implementing inclusive feedback systems that continually integrate input from diverse users could help refine information representation and reduce algorithmic distortions. Only through a shared and transparent commitment can we ensure the ethical use of AI and promote a more equitable digital society.
* The following prompt was used on ChatGPT Plus to generate the cover:
Generate a 16:9 image according to the aesthetic canons of Studio Ghibli following these guidelines:
Scene:
A young neo-fascist hacker is deeply immersed in work, sitting in his dimly lit room, focused on editing an image of an idyllic 1950s family on his computer screen. He wears a black bomber jacket, paratrooper boots, and has a shaved head. His posture is hunched, with his hands rapidly moving between the keyboard and mouse. His face is partially visible in profile, showing intense concentration.
Environment:
The room is cluttered, with piles of energy drink cans, sheets with sketches of slogans and symbols scattered across the desk, open books with right-wing historical references, and crumpled flags in the corners. The walls are adorned with slightly crumpled neo-fascist posters featuring subtle ideological signs (without explicit symbols like swastikas or Celtic crosses).
Lighting:
The setting is nighttime, with the main light source being the computer screen, creating a contrast between the cold glow of the monitor and a desk lamp with warm light, illuminating part of the room and adding depth to the image.
Computer Screen:
It is evident that the hacker is digitally manipulating the 1950s family photo. Some elements are already edited, while other parts still show the editing process, suggesting the creation of false propaganda material for an electoral campaign.
Framing:
The angle is neutral, and the subject is shot from a slightly wide three-quarter back view, allowing a glimpse of not only his work but also the messy background of the room and the objects that tell his story.
[1] https://www.biastest.ch (31.03.2025)
[2] Geschichte der Gegenwart, Echte Emotionen. Generative KI und rechte Weltbilder: https://geschichtedergegenwart.ch/echte-emotionen-generative-ki-und-rechte-weltbilder/ (31.03.2025)
[3] I bias e gli stereotipi nei generatori di immagini prodotte con l’AI: https://www.insidemarketing.it/pregiudizi-bias-immagini-generate-con-intelligenza-artificiale/ (19.06.2023); These new tools let you see for yourself how biased AI image models are: https://www.technologyreview.com/2023/03/22/1070167/these-news-tool-let-you-see-for-yourself-how-biased-ai-image-models-are/ (22.03.2023)
[4] AlgorithmWatch, Wie Bildgeneratoren die Welt zeigen: Stereotypen statt Vielfalt: https://algorithmwatch.ch/de/wie-bildgeneratoren-die-welt-zeigen-stereotypen-statt-vielfalt/ (31.03.2025)
[5] These new tools let you see for yourself how biased AI image models are: https://www.technologyreview.com/2023/03/22/1070167/these-news-tool-let-you-see-for-yourself-how-biased-ai-image-models-are/ (22.03.2023)
[6] Welt.de, Swifties als Trump-Fans, Deutsche in Ketten – die große Manipulation im Wahlkampf: https://www.welt.de/253407380 (16.09.2024)
[7] Institute for Strategic Dialogue (ISD), Generative KI und die deutsche extreme Rechte: Narrative, Taktiken und digitale Strategien: https://isdgermany.org/generative-ki-und-die-deutsche-extreme-rechte-narrative-taktiken-und-digitale-strategien/ (20.02.2025)
[8] Velia Alvich, In Europa l’estrema destra sta generando immagini con l’IA per promuovere politiche anti-immigrazione: https://www.corriere.it/tecnologia/cards/in-europa-l-estrema-destra-sta-generando-immagini-con-l-ia-per-promuovere-politiche-anti-immigrazione/lestrema-destra-e-lintelligenza-artificiale-generativa_principale.shtml (02.12.2024)
[9] Tribes.hypotheses.org, https://tribes.hypotheses.org/381: https://tribes.hypotheses.org/381 (22.01.2018)
[10] Insidemarketing.it, I deepfake e i pericoli della disinformazione nell’era dell’intelligenza artificiale: https://www.insidemarketing.it/deepfake-e-disinformazione-alcuni-esempi (16.05.2023); Agenda Digitale, IA e disinformazione: un manuale per le aziende contro le fake news: https://www.agendadigitale.eu/cultura-digitale/ia-e-disinformazione-un-manuale-per-le-aziende-contro-le-fake-news/ (29.01.2025)
[11] Altalex, L’intelligenza artificiale nella campagna elettorale: tra buoni propositi e deepfake: https://www.altalex.com/documents/news/2024/06/06/intelligenza-artificiale-campagna-elettorale-buoni-propositi-deepfake (06.06.2024); Internazionale, Come l’intelligenza artificiale può influenzare le elezioni europee: https://www.internazionale.it/notizie/2024/06/03/intelligenza-artificiale-elezioni-europee (03.06.2024)
[12] Prisma srl, Metodologie per l’identificazione delle immagini AI-generated: https://www.prismacompany.it/identificazione-delle-immagini-generate-da-intelligenza-artificiale-metodologie-e-strategie-di-verifica/ (27.032025)
[13] Commissione Europea, The EU Artificial Intelligence Act: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai (31.03.2025)
OpenEdition schlägt Ihnen vor, diesen Beitrag wie folgt zu zitieren:
Massimiliano Livi (31. März 2025). The Role of Artificial Intelligence in Shaping Collective Imagination and Reinforcing Cultural and Social Bias. TrIBES. Abgerufen am 19. Januar 2026 von https://doi.org/10.58079/13ljx
