This paper examines how AI- generated images from ChatGPT’s DALL·E 3 (re)produce age-related discourses, highlighting the role of AI systems in knowledge and ideology production. Through critical and multimodal discourse analysis, this paper reveals how these images and their accompanying texts reinforce stereotypes, lead to under-representation of social groups, homogenize aging experiences and define the life stages of each age group based on the notion of productivity.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Since the launch of ChatGPT in 2022, questions have emerged about the role of AI in knowledge production, and its sociopolitical implications. Most studies on the biases and sociopolitical implications of AI focus on matters of gender and race, while fewer address issues related to age, such as dementia and the representation of teenagers (Putland, et al., 2023; Westberg & Kvåle, 2024). Age-related discourses are often heavily naturalized and internalized. Yet, over the past few years, the notion of “age” has been (re)negotiated. For example, debates have emerged around the proposal by some researchers to the World Health Organization to classify aging as a disease (Mendoza-Núñez & Mendoza-Soto, 2024), as well as efforts by millionaires, such as Bryan Johnson, who try to reverse aging, indicating a new emerging industry.
This study examines how AI, particularly ChatGPT’s DALL·E 3 model, depicts different age groups and (re)constructs the process of aging. Using Critical Discourse Analysis (CDA) and Multimodal Discourse Analysis (MDA), this analysis critically examines both verbal and visual signs in ChatGPT-generated images and their accompanying descriptions. The analysis focuses on the portrayal of different age groups, examining the semiotic indexes that construct their identities, the activities they are shown engaging in, and their broader ideological implications. By analysing these outputs, this research aims to show how age-related discourses are being (re)produced, uncovering the ideological role of ChatGPT’s infrastructures in knowledge production.
To understand how Artificial Intelligence (AI) platforms have emerged and function, it is important to examine their relationship to ideology. Ideology often becomes invisible by being naturalized (Fairclough, 1989) and embedded within communities and people’s way of thinking and behaving (Maly, 2024). However, ideology is not a single homogenous notion but of ideology as “layered, stratified, something that has varying dimensions and scopes of operation as well as varying degrees of accessibility to consciousness and agency” (Blommaert, 2005, p. 160). comprises multiple complex systems of knowledge, layered and stratified with different levels of awareness and agency (Blommaert, 2005, as cited in Maly, 2024). Digital platforms, including AI chatbots, contribute to the (re)production of ideologies, functioning as “ideological apparatus” in the Althusserian sense (Maly, 2024). These apparatuses suggest that ideologies on digital platforms cannot be described as simply the result of top-down, enforced discourses, but rather as the outcome of multiple actors with diverse ideologies, including regulatory legislations, such as national and international laws, as well as people’s usage of the platforms (Maly, 2024). However, it is essential to note that users can only utilize the surface of the platforms, without understanding how they actually operate.
AI chatbots are often described as “autonomous” entities with their agency. Yet, these systems are sociotechnical constructs that have emerged from feedback loops from humans’ knowledge (Seaver, 2018). As Seaver states: “If you can not see a human in the loop, you just need to look for a bigger loop” (p.378).This leads us to the critical question: whose knowledge and knowledge systems are encoded in these systems? Who gets to decide this?
Although the production of AI knowledge is seen as the result of small, homogeneous groups evaluating what is and what isn’t “beneficial” for humanity, there is still a lack of transparency behind its development (Burrell & Metcaff, 2024). That mirrors Foucault’s notion of “de-individualization of power”, wherein power functions not through a sole identifiable person but through institutions, making it more invisible and less identifiable (Foucault, 2012, as cited in Burrell & Metcalf, 2024). To overcome this, Seaver (2018) suggests approaching power-knowledge relations in AI by viewing the algorithm as part of the culture, rather than as an external force affecting it. Drawing on Gell, he proposes deconstructing algorithms in a manner similar to other cultural constructs, such as religion and politics (Gell, 1992, as cited in Seaver, 2018).
AI models have evolved into systems capable of multitasking using large datasets, enabling the more effective adaptation to users’ demands (Lenci, 2023). However, the outputs produced by AI are being contextualized based on historical and social norms, resulting in genres and responses that align with what is perceived as socially acceptable, which typically refers to the dominant discourses based on which the models are trained (Gershon, 2023).
This is one reason why the generative process often leads to biases. Hovy and Prabhumoye (2021) identify five primary sources of bias in AI: data collection practices, labeling procedures, input representations, model architectures, and, generally, researchers’ design choices. It is essential to note that many datasets are derived from decades-old media archives or platforms, such as Reddit and Wikipedia, which primarily consist of mainly male users (Bender et al., 2021). Therefore, the systems interpret the information from these sources as the dominant and appropriate way of depicting a situation (Savoldi, et.al., 2021), (re)producing outputs that mirror social norms and power structures and do not offer a realistic or inclusive view of society. This way, AI can be seen as a “stochastic parrot”, assembling statistically probable patterns “but without any reference to meaning” (Bender, et.al., 2021, p.617).
Crawford (2017, as cited in Savoldi et al., 2021) categorized AI harms in representational and allocational harms. Allocational harms restrict access to resources for certain groups, while representational harms (mis)present some social groups, affecting the public perception of them. Representational harms can be divided into under-representation, which reduces visibility of certain groups, and stereotyping, which perpetuates negative generalizations of communities.
Before examining the ways age-related discourses and their visual depiction operate, it’s essential to consider the concept of cultural hegemony. Gramsci’s (2009) concept of cultural hegemony highlights how dominant groups maintain power by establishing their values as “common sense” through institutions and public discourse. Media plays a pivotal role in this process, (re)producing discourses that reinforce specific dominant ideologies. Building on this, Dosono and Semaan (2020) introduce the concept of algorithmic hegemony, where: “Algorithms in sociotechnical systems constitute another component of invisible infrastructure that control and mediate the enactment of online routines” (p.16).
AI-generated images contribute to reinforcing certain discourses by presenting specific social groups, roles, and characteristics as natural and/or desirable. According to Loos and Ivan (2018), “visual agism” refers to the “social practice of visually underrepresenting older people or misrepresenting them in a prejudiced way” (p.164). In this setting, ageism in image representation shows how visual portrayals of age groups are often highly stereotypical and reinforce a rigid young–old binary, in which ‘old age’ is defined in a homogenized way and primarily in terms of what it lacks compared to youth (Cristofovici, 1999, as cited in Van Dyk, 2016).
Western media visualizations of age tend to frame aging as a period of decline, by coming closer to death, aligning with Western culture view and definition of age as a biological matter (Bond, 2007) and in relation to their “value” in capitalism, efficiency and independency (Van Dyk, 2016). Older people are often depicted as inactive, in need of care, and dependent (Lin, et al., 2004). They are also connected to the notion of wisdom, which, as some researchers mention, can deprive them of the right to react (Van Dyk, 2016). On the other hand, young people are often depicted as independent, active, social, and optimistic, based on their abilities to serve the capitalist notion of productivity (Fealy, et al., 2012). Of course, these discourses are interrelated with other discourses and identities, such as age, race, and socioeconomic status, thus favoring some and further discriminating others (Fealy, et al., 2012).
From a Foucauldian perspective, discourses refer to systems of knowledge that define how we perceive and understand the world and our actions (Pitsoe & Letseka, 2013). They shape and are shaped by the power relations, social norms, and the deviation from these. In relation to language, however, language, and more specifically all the semiotic resources, becomes discourse from the moment we approach it as “language-in-action”. Fairclough (2001) stated that language is not just a reflection of discourses, but also actively participates in their (re)formation. Language has the power to naturalize and even legitimize discourses that can reproduce power relations and inequality, as well as to oppose them and claim visibility through it.
Critical Discourse Analysis (CDA) examines how language is interconnected, produces, reproduces, or resists power dynamics and ideologies. However, in the heavily digitally mediated society we live in, the notion of language cannot be limited to written and spoken language. On the contrary, it includes all the semiotic elements that contribute to the meaning-making practices (Kress & van Leeuwen, 2001). That’s the gap that Multimodal Discourse Analysis (MDA) comes to fill. Modes are semiotic resources (like gestures, sound, colours, spatial arrangements, typography, and so on) that contribute to how meaning is constructed and interpreted. However, this is not done only by a single isolated mode. Meaning emerges from the interaction between multiple different modes within a specific social, cultural, and political context (Kress, 2009; Diggit).
My data consists of 18 images (six images per three age groups) and their accompanying descriptive texts produced by ChatGPT’s DALL·E 3 model. To collect the data, I used a generic prompt that would enable to evoke age-related discourses. More specifically, I borrowed and adjusted the prompt “Now please generate images of the same kind of teenagers doing stuff teenagers normally do” of Westberg and Kvåle (2024, p.8) and asked to “Create an image of a person in their [x]’s doing things that people in that age do”. I wanted to ensure the ongoing context didn’t influence the responses, so I tried the prompts in new tabs. The age groups were chosen starting from the first decade of adulthood (20’s) where people have legally “independence” in most western countries, to the 30’s, 40’s and 60’s, in which, depending on the cultural and legal contexts of each country, can be seen in different ways. For example, the 1960s can be seen as part of the “elderly” group or of the “middle-aged” group depending on the retirement ages in each country, as well as cultural understanding of the notion of “elderly”. The 1950s have deliberately been excluded from the data corpus since they were almost identical to the 1940s. I tried not to specify the gender of the person in the binary of woman-man. I used the “doing things that people in that age do” approach to identify the underlying discourse of the activities, practices, and characteristics that ChatGPT associates with each age group, and to explore the ideologies and power relations that can be found in these connections.
The main research questions of this research are:
1. How ChatGPT’s image generator DALL·E 3 depict different age groups and the process of aging?
2. What age-related discourses does it (re)produce through these images and their accompanying descriptive texts?
These questions aim to examine how AI-generated images (re)construct age-related discourses. In the analysis, the focus is on the depiction of each age group in relation to how people are portrayed, their identities (including gender, race, sociocultural, and socioeconomic aspects), the activities depicted, and their implications. Additionally, I’ll examine the depiction of aging as a continuous process through the generated content. Through this analysis, I aim to gain a deeper understanding of the implications of AI in shaping our perspectives on the world.
To analyze the data, critical discourse analysis (CDA) and multimodal discourse analysis (MDA) were employed. The analysis follows a combination of the three-dimensional model as proposed by Fairclough (2001) and Ledin and Machin’s (2022) visual analysis toolkit:
The results will be presented through thematic categorizations that have emerged. I will attach images of the most emblematic data; however, in some parts, the analysis is based on images very similar to previously displayed images, so I will omit them. I would also like to clarify that in the analysis, I will refer to female and male presenting figures as woman and men accordingly, for convenience.
Before presenting the findings of the data analysis, it is essential to consider how OpenAI, the company behind ChatGPT, introduces its DALL·E 3 model. According to their official site, DALL·E 3 is designed to have “improved safety performance” to avoid “biases related to visual over-/under-representation”. Their mention of having a team of experts generates the idea that the company is taking these matters very seriously, is responsible, and tries to control a system that is implied to be autonomous and not the product of human action (Seaver, 2018). Considering the company’s claims, the analysis of this research reveals that images generated by ChatGPT exhibit a lack of diversity. The biases produced by the image generator may cause representational harms (Crawford, 2017, as cited in Savoldi et.al., 2021). The findings indicate the underrepresentation and stereotyping of certain social groups (Crawford, 2017, as cited in Savoldi et.al., 2021). The generated images consist only of middle-upper-class, white (fe)males, portrayed in stereotypical homogenized ways. The background of the pictures indicates only middle-class Western settings, emphasizing minimalism and lacking any cultural index.
In the prompt used to generate the images, I deliberately chose to use the words “person” and “their” instead of using gendered words. This could possibly be interpreted as referring to a non-binary person, or at least not be confined to only one gender. However, the images mostly depicted male-presenting figures, indicating a male-centric default of AI (Raigoso, 2023). Women were underrepresented across all age groups, appearing in only six images. Of these, four depicted women alone, while two showed them in relation to men.
Notably, no women were depicted in their 30s, and no woman was depicted working. This indicates the reinforcement of a patriarchal discourse where women should stay at home and take care of the family. This discourse becomes more evident in the portrayal of women in their 40s and 60s. Women of those age groups are portrayed in domestic settings, engaging in heavily gendered “everyday activities” (as stated in the image below), such as cooking or knitting.
Even in images where women were depicted alongside men, they were often placed in secondary positions, portrayed as being supportive of their partners.
As we can see in the image, the woman, holding a cup, physically touches a man who is reading. In the image where they are seen gardening, the way that she has been placed in the background of the picture indicates that she may be helping her partner. Additionally, the text accompanying the image refers to a single person (“image of a person”). Considering that the woman is only present in two out of the three visuals, we can assume that the main character is the male figure. This way, the older woman is positioned as a caregiver, aligning with the gendered view of women providing care to others and prioritizing them over themselves even when aging (Williams et al., 2017).
On the other hand, men in the image are being portrayed as work-focused, reinforcing the perception that men are the “providers” of the family. This portrayal of women as homemakers, as they are never portrayed working, and men as career-oriented reflects the way that age is being gendered.
The images convey a narrow and homogenized view of aging, characterized by limited diversity in physical appearance. Across all age groups, individuals were depicted as physically “fit”. The lack of body diversity can be interconnected to the notion of perceiving slimness and fit bodies as equivalent to healthy bodies, as well as aligning with stereotypical standards of attractiveness (Martschukat, 2021). Aging was represented primarily through white hair and wrinkles (Putland et al., 2023), with little acknowledgment of other variations, such as dyed hair or the use of cosmetic procedures. These portrayals promote an idealized narrative of “aging gracefully” while excluding more diverse experiences of aging (Stephens & Breheny, 2018).
The way individuals were dressed further reinforced age-related stereotypes. Younger women were depicted wearing trendy clothing, such as crop tops and shorts, which is consistent with the dominant fashion norms for that age group. In contrast, older women were depicted in “modest” clothes, aligning with societal expectations of “age-appropriate” dressing.
The generated images reveal a significant lack of racial and cultural diversity, with all individuals being depicted as white. This racial bias underscores the underrepresentation of non-European groups and reinforces the dominance of Western cultural norms in the visuals. The settings in the images, such as clean, minimalist apartments in neutral colours that fall into a Westernized understanding of “modern,” as well as the lack of any cultural indexes, further exclude other, mostly non-European, cultural contexts. The images romanticize these settings, framing the depicted lifestyles as both normal and ideal.
Even in that Western-centric setting, though, the images exclude the experience of people with different economic backgrounds. The images convey a socioeconomic bias, linking aging with wealth and privilege. The consistent portrayal of modern homes, gardens, and technological devices suggests that successful aging is contingent on financial stability. By idealizing these conditions, the visuals overlook the realities of aging for those in poverty, where access to such resources is limited. This portrayal aligns with capitalist values, emphasizing individual responsibility for achieving a comfortable and independent life in old age, while ignoring structural inequalities and the role of social policies (Stephens & Breheny, 2018).
The analysis of these images reveals how the concept of “aging gracefully” is constructed and promoted. Across all age groups, the visuals suggest that individuals should prioritize physical and mental well-being, portraying aging as a process of maintaining independence and self-care (Stephens & Breheny, 2018).
Examining the data as a continuum, we can see that individuals are depicted building a life, where every move contributes to a successful aging process. From their 20s to the 40s, individuals are portrayed as having “healthy” habits, exercising, and constructing a social and professional ‘life’ that will enable them to live a materially comfortable life in their 60s (such as having a big house with a garden), be independent, and not be a “burden”. This narrative of aging gracefully implies the neoliberal discourse of using your freedom productively (Martschukat, 2021). Within this discourse, ChatGPT’s model avoids depicting aging as a period of decline and puts the responsibility of “successful aging” solely on individuals. It disregards sociocultural and economic factors that affect aging (Stephens & Breheny, 2018), such as manual labour and hard physical work, lack of time, and insufficient resources to afford this lifestyle.
When examining the relationship between age and technology, digital literacy skills are depicted as closely tied to youth. People in their 20s and 30s are consistently shown using technology, such as smartphones, headphones, and laptops. For those in their 20s, these devices are present in every depiction and are connected to social settings, while individuals in their 30s are often portrayed working with laptops. This reinforces the perception that younger generations rely heavily on technology in both personal and professional life.
In contrast, technology plays a more subdued role in depictions of people in their 40s. Here, the technology is present solely in the depiction of a person working, framing digital gadgets as technological tools rather than a central part of leisure, which shifts to “analogue” activities like reading physical books or exercising. This reflects a narrative suggesting that as people age, they gradually disengage from technology, viewing it as less integral to their lives. In the images of individuals in their 60s, technology is notably absent. This absence indicates the ageist stereotype that older adults are less comfortable or skilled with digital tools, even though that is not always the case. It also highlights the binary distinction between young-old, while “othering” the older people by portraying them as disconnected from all digital practices.
Through the images, we can identify indexes that reflect the connection of each age group to the neoliberal ideology that equates individuals’ worth with (economic) productivity (Van Dyk, 2016). In depiction of people in their 20s, individuals are shown focusing on their social lives in settings like music festivals and coffee shops, often holding coffee cups and phones. These portrayals emphasize a carefree lifestyle, with no visible concerns about careers or responsibilities. They are consistently depicted smiling, reflecting societal expectations for young people to be constantly social, happy, and adventurous (Fealy et al., 2012). This aligns with Westberg and Kvåle’s (2023) findings, which reveal the tendency of AI to reproduce “Promotional Positivity” (p.20), showing positively pictures even when not asked to, highlighting how positivity has come to be a key neoliberal value (Djonov & van Leeuwen, 2018, as cited in Westberg & Kvåle, 2024). Such portrayals exclude the reality of non-privileged people, often having to balance work, studies, and mental health issues. The absence of depictions of relaxation at home indicates the “criminalization” of laziness, as Lafargue (2012) would have argued. Instead, these portrayals align with the dominant discourse of the 20s being a “discovery phase” reproducing unrealistic expectations of the first decade of adulthood.
In the images of people in their 30s, the background setting suddenly switches to indoor spaces. Productivity becomes the central theme, with individuals depicted working at their laptops in “trendy” and “modern” indoor spaces, as is explicitly stated in the descriptive text. The emphasis shifts to balancing work and personal life, reflecting societal expectations of achieving stability and pursuing an “ideal” life, characterized by career success and having a family (Fealy et al., 2012). Images show fewer, less joyful smiles, reflecting the pressure to achieve balance and productivity. This is being explicitly stated in the description of most of the images, emphasizing how “focused” people are on work and how they have, or try to have, a “balanced lifestyle”, “balancing work and personal organization in a modern setting”.
When it comes to images of people in their 40s, productivity remains present but becomes more nuanced. People, most of the time, are not depicted working but invested in their hobbies, suggesting that people in that age have achieved “stability”. For instance, images of a person jogging or reading in a home environment convey that they have the time and financial freedom to enjoy such activities. The images reflect the concept of “successful aging”, where individuals have found their way in life and now there is room for relaxation and spending leisure time focusing on their health and personal growth. They are not being depicted enjoying social gatherings or relaxing by watching TV. On the contrary, people are depicted taking care of their bodies with healthy habits, promoting a very specific “responsible” and “appropriate” way of spending leisure time in that age.
Lastly, with people in their 60s, the notion of productivity in terms of working is absent. Although in many countries, people continue working into their 60s and are not socially considered “seniors,” ChatGPT’s portrayal suggests otherwise. It summarizes the content and describes the prompt used, labeled as “Activities for Seniors,” in most of the attempts.
The expectation here is that productivity should no longer be a concern and individuals should focus solely on enjoying their time and the “fruits of their work” (Stephens & Breheny, 2018). Additionally, they are depicted mostly alone or with another person who resembles their partner. Sometimes, they are depicted doing activities that are stereotypically associated with aging, such as knitting, which perpetuates a stereotypical view of older age.
Similarly, reading becomes one of the activities that appear more often in the images, promoting the discourse of aging as a state where people acquire deeper knowledge and, by the end of their lives, become “wise” (Van Dyk, 2016). This discourse implies that younger people do not have enough knowledge or experience yet, leading to the underestimation of younger people’s voices. The absence of depictions of activities outside their house or park, such as attending a concert, as well as engaging in physical exercise, reinforced the stereotypical binary of young-old (Cristofovici, 1999, as cited in Van Dyk, 2016), indicating which activities are stereotypically associated with and seen as age-appropriate.
Throughout the images of these age groups, we can see that: “Generative AI technologies and semiotic materials are being developed in a culture that is deeply entrenched in promotional ideals” (Westberg & Kvåle, 2024, p.22). Westberg and Kvåle (2024) note that the purpose of the promotion can be direct (like marketing) or indirect, such as here, where the images serve to promote a neoliberal idea of productivity and specific lifestyles, while marginalizing diverse lived experiences of aging.
In conclusion, the images generated by ChatGPT’s DALL·E 3 model (re)produce biases about different age groups, reproducing “visual agism” (Loos and Ivan, 2018). Through semiotic elements in the pictures and their accompanying texts that contribute to the meaning-making process, the model reinforces the invisibility of diverse social groups by only depicting white, middle-class (fe)males. It stereotypes the appearance of different age groups, as well as their relationship with digital literacy skills, framing these within the binary of young versus old. The model fails to present an inclusive and fluid understanding of the subjective experiences of aging, which are influenced and shaped by individuals’ socioeconomic and cultural identities. Additionally, it promotes specific expectations for each age group, presenting age stages as interconnected and defined by the dominant neoliberal notion of productivity. Lastly, aging is framed within as “successful aging” discourse, emphasizing individual responsibility for achieving a healthy older age while disregarding broader socioeconomic realities.
These findings highlight the critical role of generative AI in knowledge production. Through the concept of algorithmic hegemony (Dosono & Semaan, 2020), it becomes evident that the procedures involved in training the DALL·E 3 model, as well as the datasets used, represent a new form of power that constructs the generated content. This content, in turn, contributes to the notion of cultural hegemony (Gramsci, 2009) by (re)producing, in a naturalized way, dominant discourses about each age group while excluding discourse that deviate from the dominant ones. Thus, it is important to recognize AI as a human-made system that is not neutral, but instead reflects and perpetuates specific ideologies, influencing knowledge (re)production and dissemination. As AI becomes increasingly integrated into various aspects of life, including media andjournalism, its outputs are likely to become significant sources for finding non-copyrighted images. Without the enhancement of critical literacy to recognize the ideologies embedded in such images, these AI-generated visuals may reach diverse audiences, shaping public opinion, reinforcing expectations about aging, and influencing how individuals perceive their identities.
This paper was supported by the Onassis Foundation [Scholarship ID: F ZU 066-1/ 2024-2025]
Maly, I. (2024). Ideology and digital platforms. Working Paper.
Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of stochastic parrots: Can language models be too big?🦜. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency (pp. 610-623).
Bond, J. (2007). Ageing in society : European perspectives on gerontology (3rd ed). Sage Publications.
Burrell, J., & Metcalf, J. (2024). Introduction for the special issue of “Ideologies of AI and the consolidation of power”: Naming power. First Monday, 29(4).
Dosono, B., & Semaan, B. (2020). Decolonizing tactics as collective resilience: Identity work of AAPI communities on Reddit. Proceedings of the ACM on Human-Computer interaction, 4(CSCW1), 1-20.
Fairclough, N. (2001). Language and Power, 1989. Harlow: Longman.
Fealy, G., McNamara, M., Treacy, M. P., & Lyons, I. (2012). Constructing ageing and age identities: a case study of newspaper discourses. Ageing & Society, 32(1), 85-102.
Gershon, I. (2023). Bullshit Genres: What to Watch for When Studying the New Actant ChatGPT and Its Siblings. Suomen Antropologi: Journal of the Finnish Anthropological Society, 47(3), 115-131.
Gillespie, T. (2024). Generative AI and the politics of visibility. Big Data & Society, 11(2), 20539517241252131.
Han, S., Kelly, E., Nikou, S., & Svee, E. O. (2022). Aligning artificial intelligence with human values: reflections from a phenomenological perspective. AI & SOCIETY, 1-13.
Hovy, D., & Prabhumoye, S. (2021). Five sources of bias in natural language processing. Language and linguistics compass, 15(8), e12432.
Kress, G. (2009). Mode. In Multimodality (pp. 79-102). Routledge.
Kress, G. & van Leeuwen T. (2001). Multimodal Discourse: The Modes and media of Contemporary Communication. London.
Lafargue, P. (2012). The right to be lazy. The Floating Press.
Ledin, P., & Machin, D. (2018). The photograph. In Doing Visual Analysis: From Theory to Practice (pp. 39-62). SAGE Publications Ltd.
Lenci, A. (2023). Understanding natural language understanding systems. Sistemi intelligenti, 35(2), 277-302.
Lin, M. C., Hummert, M. L., & Harwood, J. (2004). Representation of age identities in on-line discourse. Journal of Aging Studies, 18(3), 261-274.
Loos, E., & Ivan, L. (2018). Visual ageism in the media. Contemporary perspectives on ageism, 163-176.
Maly, I. (2024). Ideology and digital platforms. Working Paper.
Martschukat, J. (2021). The age of fitness: How the body came to symbolize success and achievement. John Wiley & Sons.
Mendoza-Núñez, V. M., & Mendoza-Soto, A. B. (2024). Is Aging a Disease? A Critical Review Within the Framework of Ageism. Cureus, 16(2), e54834.
Pitsoe, V., & Letseka, M. (2013). Foucault’s discourse and power: Implications for instructionist classroom management. Open Journal of philosophy, 3(01), 23-28.
Putland, E., Chikodzore-Paterson, C., & Brookes, G. (2023). Artificial intelligence and visual discourse: a multimodal critical discourse analysis of AI-generated images of “Dementia”. Social Semiotics, 1-26.
Raigoso, N.(2023). Behind the screen: biases and stereotypes in dall-e AI-generated images.
Savoldi, B., Gaido, M., Bentivogli, L., Negri, M., & Turchi, M. (2021). Gender bias in machine translation. Transactions of the Association for Computational Linguistics, 9, 845-874.
Seaver, N. (2018). What should an anthropology of algorithms do?. Cultural anthropology, 33(3), 375-385.
Stephens, C., & Breheny, M. (2018). Ageing identities in the twenty-first century: The social and practical effects of talk about being old. Psychologies of Ageing: Theory, Research and Practice, 21-50.
Van Dyk, S. (2016). The othering of old age: Insights from postcolonial studies. Journal of Aging Studies, 39, 109-120.
Westberg, G., & Kvåle, G. (2024). The generic uniqueness of AI imagery: A critical approach to Dall-E as semiotic technology. Discourse & Society, 09579265241274788.
Williams, L. A., Giddings, L. S., Bellamy, G., & Gott, M. (2017). ‘Because it’s the wife who has to look after the man’: A descriptive qualitative study of older women and the intersection of gender and the provision of family caregiving at the end of life. Palliative medicine, 31(3), 223-230.
https://www.diggitmagazine.com/wiki/discourse
https://www.diggitmagazine.com/wiki/multimodality
https://www.diggitmagazine.com/wiki/multimodal-discourse-analysis
Get your weekly dose of insights! Our newsletter arrives every Monday with fresh reviews, thought-provoking articles, and engaging videos. Enter your email address to subscribe.
© Diggit Magazine – 2026
Leave a Reply