The text below is a brief excerpt from my final thesis in sociology under the title "‘AI’ in situated action. Unlocking the role of UX and UI design in enabling ‘AI’ in everyday life". (Based on empirical data from a research study I conducted.)

The continuous reiteration of ‘AI’ as an autonomous technology constitutes ‘AI’ as a fact  

Practices of naturalization and decontextualization, through which facts are constituted, have been widely examined in Science and Technology Studies. The continuous reiteration of ‘AI’ as a self-evident or autonomous technology does just that. It constitutes ‘AI’ as a fact (Suchman 2023). In doing so, certain aspects, such as design decisions and construction processes, or even related technical infrastructures, are being made invisible (Pentenrieder/Weber 2020). 
[...] 
Ensuring the transparency of these practices and interfaces is an ambitious project.
[...] 

The attribution of status to AI is made possible by collective action

A key insight from a more recent study on social status and the moral acceptance of AI is that the higher the status of an organization using AI, the greater the acceptance of that AI as a moral actor (Schenk et al. 2024:1002). If so, does this imply that the status attributed to AI also influences its everyday use? I argue that the attribution of status and its translation (see Latour 1996; Callon 2020) to AI is made possible by collective action. UX and UI designers are involved in this. 
[...] 
UX, UI, and Conversational designers are responsible for the creation of interfaces and the framing of situations in which human actors interact with products or services. This creation is not neutral and does not occur in a vacuum; rather, it is context-specific and culturally coded. Therefore, technologies also reflect a particular distribution of power and accountability (Pentenrieder/Weber 2020; Goffman 1963; Foucault 1997 [1970]).
[...]
Those who are actively involved in shaping situations of human-AI interactions consequently have power and responsibility. 
[...] 
This underscores: Designers are not only humans in the loop, but also moral actors. Their attitudes, practices, and knowledge are of interest.
[...] 

What is meant by situation and situated action? 

In sociology, situations are either localized at the level of interactions (between those present) or at the level of phenomena (Miko-Schefzig 2022). For my approach, it makes sense to link these two levels and consider them as a continuum. In the context of UX and UI design, interactions are central. However, these interactions form the foundation for analyzing how AI is enabled as a social phenomenon. 
[...]
Simultaneously: no one enters a situation (of interaction) as a literally blank sheet of paper, interpreting it solely from within themselves, i.e., based on the situation. Everything situational is also situated, i.e., already socially shaped (see Goffman 1963, 2009; Reichertz 2013; Miko-Schefzig 2022). In light of this, Suchman (1985: 35) uses “the term situated action. [It] underscores the fact that the course of action depends in essential ways upon the action's circumstances”. 
[...]
Therefore, I define specific projects in which AI is integrated as my situation of situated action. 
[...]
I identified three categories of UX and UI design situations in which AI was incorporated: (1) UX and UI design for AI models themselves. For example, the design of the interface and the user experience of the Chatbot Pi. (2) Integrating AI into the design process when conceptualizing products and services. For example, designers might use ChatGPT to identify patterns in research data, or they might use Vibe coding for prototyping. And (3) incorporating AI models into final products or services as part of the UX and UI design of these products and services. One example would be a customer service chatbot for a financial institution. 
[...] 
Apparently, designing with 'AI' is often involved in designing for AI. It is also involved in designing products in which AI is incorporated. 
[...] 

Furthermore, context matters

For example, it makes a difference to the situation of situated action whether the aim is to incorporate an AI model into an app for vegan nutrition that is designed to motivate people to consume a variety of foods, or whether it is an app that provides (medical) advice for cancer patients.
[...]  
When analyzing a situation in regard to enablers, it also might be beneficial to start by identifying the conditions that make it impossible (see Abend 2020). The respondents' statements repeatedly addressed the extent to which, and the properties of AI that make the integration of AI into UX and UI products and services challenging or impossible. For example: "The biggest limitation, and I don't know if the problem will ever be solved, is the hallucination problem". Another respondent expressed the same concern and added considerations regarding a possible loss of status for the company. "There are a lot of cases in the industry where you can hear about when chatbots have gone rogue, and they have a massive potential for reputational damage".
[...] 
Then there are legal, cultural, or ethical concerns that might lead to a lack of acceptance or conflicts between stakeholders. This, in turn, can hinder technological progress or the integration of AI into everyday life. For example, attitudes toward data security and privacy.
[...] 

Everything that appears to be situational is actually situated 

(1) The situatedness of the situation is the immediate experience of interaction in time and place. In an interaction between humans and AI, such as a chatbot, this means, for example: What wording does the AI use, or how does the voice sound? Is the chatbot hallucinating? etc. But, and this is what is unique, even if it is an interaction with a generative chatbot, everything that appears to be situational is actually situated. Because (2) in regard to AI, everything situational is contextualized in advance, for example, by the training data. Or in UX and UI design, the bot has been given a backstory or pre-scripted answers or guardrails, defining what the chatbot can and cannot say, and so on. Furthermore, AI is usually incorporated into a product and thus additionally situated insofar as it is assigned an authorship (and its status). For example, it is a customer service bot for Bank A. Even if the chatbot itself is the product, such as ChatGPT, it is still linked to the status of OpenAI. UX and UI design, as part of the sociotechnical system, define the specific conditions of these interactions.
[...] 

Interactions are at the core  

As I argued initially, UX, UI, and C-UI designers are responsible for the creation of interfaces and the framing of situations in which human actors interact with products and services. Consequently, interactions are at the core. Generally speaking, in the field of UX and UI design, AI is incorporated in everyday products that require interactivity. In this respect, AI (or its technological properties) is enabled to meet this requirement. However, as my research emphasizes, the specific integration is invariably context-specific, relational, and culturally coded. It is negotiated in the process, among other things, as part of a collective action. While human actors interact with AI, the meaning of AI itself is negotiated. This has a (reciprocal) influence on the extent to which AI is enabled in the UX and UI of everyday products and services. This is due to the expectations and needs of users.

If you’d like to find out more (about my findings) or discuss any specific points, please don’t hesitate to get in touch!