Users are "architects" of their environments, as Kirsh put it. For example, users may change the settings to turn on or off a function or change the way it behaves. They also choose the applications they use. Such tailoring behaviors are not explained by Norman's intention–action–response–interpretation–evaluation cycle.

The How was School today. . .? concept was developed iteratively with two children with CP and the help of school staff. This AAC system was designed to use context to facilitate the creation of personal narratives [75]. The authors called their approach "data-to-text": The idea was to add sensors to the environment and the wheelchair of a user.

The families also exaggerated articulation, a phenomenon known as hyperarticulation. The paradoxical effect of hyperarticulation is that despite trying to improve understanding, it can make speech recognition worse.

Their study exposed the limited nature of contemporary speech interaction from a conversational perspective. Although breakdowns were not that frequent—one occurred every four hours of use—they disrupted regular use and often required joint effort to overcome.

According to Scholtz [745], the two gulfs manifest differently in the different roles a user may have when interacting with a robot: Supervisor, Operator, Peer, Bystander.

The results of the studies show that a small set of responses (3–9) may be sufficient to generate an idea of what a prompt can do; the computation of more responses might just waste time. The results also show that the SUBJECT would sometimes get lost in the STYLE; some prompts inadvertently led to grotesque or inappropriate images.

The challenge here is sometimes described as prompt engineering—the search for prompts that give the output the user finds adequate for the task.

Liu and Chilton [488] noted that interaction with such models faces a dilemma. While it is possible to input anything as a prompt to such models, users must "engage in bruteforce trial and error with the text prompt when the result quality is poor."

One thing that is missing is an account of how beliefs about the computer are formed and updated and how they drive action specification. The current understanding is that users form internal models that predict how their actions produce perceived outputs, and they learn to minimize prediction errors.

both the computer and the human participate in establishing a shared context. The computer does not simply receive a message; it also communicates the effects of that message.

Kirsh argued that we are not just passively reacting to computer-generated options. If we look at interaction at a higher level, beyond a single action, we see that users are also actively influencing their environments. Users are "architects" of their environments, as Kirsh put it.

Code-switching refers to a switch in language to match the capabilities of the communication partner... Such differences are important because depending on the communication context, people will have different expectations and styles they use in dialogue with a computer.