We also identified the factors that are critical to older adults but did not appear in the existing models. Finally, we applied the existing vocabulary to our model to comply with the conventional terms in the field.

Again following grounded theory practices from [33], we compared the model that emerged from our data with existing theoretical models of technology acceptance to determine differences and similarities between them.

Triangulating the empirical findings from our preliminary results with the existing theoretical models, we proposed an extension of the existing theoretical models that explains the technology acceptance behavior of our participants who were aged 60 or over.

Consolidating our preliminary findings with the existing models, we propose an extended technology acceptance model for older adults illustrated in Figure 3. Extending to the predecessor theories, our tentative model introduces the perceived effort of learning a new technology as an obstacle for older adults' technology acceptance, which has not been reported in any studies of younger adults' technology acceptance.

Using TAM, UTAUT, and several other works as theoretical underpinning, Renaud and Biljon proposed a model to explain older adults' mobile phone adoption.

Although many researchers have sought to understand and predict technology acceptance behavior, there has been relatively less effort to build a theoretical model for older adults, with one exception (STAM).

Extending the original TAM and consolidating the constructs of several other existing models, Venkatesh et al. proposed the Unified Theory of Acceptance and Use of Technology (UTAUT) [37].

Azjen's theory of planned behavior [1, 2] posits that a specific behavior is the result of an intention to carry it out, and that intention is determined by attitudes, norms, and the perception of control over the behavior. Drawing upon this theory of planned behavior, Davis et al. developed the technology acceptance model (TAM) [10].

Then, by triangulating our empirical findings with existing theoretical models from the literature, we found out that the existing models of technology adoption require new theory components to be able to describe technology adoption processes of our participants.

Technology acceptance has been widely studied, and several models have been proposed and tested [10, 37]. However, the HCI literature lacks a comprehensive explanation of technology acceptance among older adults.

The beauty of the GP-TSM technique lies in its simplicity: at its core, all GP-TSM does is change the visual saliency of words by adjusting their opacity. This preserves the integrity of the original text and minimizes "ergonomic obtrusiveness" [100] while providing readers with a form of "contextual cuing" to arm them with "incidental knowledge about global context", which they can harness to better assign visual attention and memory when reading [40].

Furthermore, according to Stevens's power law, people perceive changes in gray scale not linearly, but rather by a factor of approximately 0.5 [71]. For instance, a threefold increase in opacity might only be perceived as 1.5 times more significant, further complicating the differentiation of levels.

This sequence resonates with efficient content absorption strategies highlighted in speed reading literature, where readers first capture the gist and then delve deeper [1, 63]. The interface, therefore, may inadvertently facilitate this structured, layered reading approach, which might explain the improvement in reading efficiency and comprehension.

We adopt the term "saliency" based on its definition (a "bottom-up, stimulus-driven perceptual quality which makes some items stand out from their neighbors") [42], and its use in augmented reality [85, 88], computer vision [17, 55], and cognitive science [37, 56].

Modulating text saliency is a widely studied aspect of textual information representation. This technique modifies the visual attributes of text to promote words of interest and guide readers' attention, making pertinent information more perceptible and thereby enhancing comprehension and the user experience [12, 42].

compressive summarization aims to select the shortest subsequence of words within a sentence that yields an informative and grammatical sentence [64]. This framework allows for a more concise representation of the original content while retaining the essence of its meaning.

Given the cognitive effort reading requires, readers frequently resort to skimming, which is a rapid, selective, and non-linear form of reading [2]. Eye tracking studies [30, 74] validate that such behavior is extremely common. However, multiple studies have suggested a significant trade-off between reading speed and comprehension [65, 66, 76, 87].

Specifically, automated summarization methods can introduce multiple types of errors: "crimes" of omission, hallucination, and misrepresentation.

Automated text summarization techniques, including but not limited to crowd-powered systems [10], prompting large language models (LLMs) [105], and other AI technologies, can address a subset of these difficulties, i.e., the resulting text may be shorter, with simpler sentence structures and fewer unusual words [62]. However, unless there is information within the original document that is truly redundant, the result is a lossy representation of the original document, regardless of whether the process is abstractive or extractive.

Established theories of human cognition describe how exposure to variation and consistency within prescribed structures can help people more robustly form mental models of a phenomenon, e.g., how an LLM behaves. Specifically, in line with Variation Theory [35], the features we instantiate identify patterns of consistency (Figure 1d, "Exact Matches"), variation (Figure 1c, "Unique Words"), or both (Figures 1a, 1b, "Positional Diction Clustering (PDC)"—a novel algorithm we introduce in this paper). In line with Analogical Learning Theory [13], PDC highlights analogous text across LLM responses, i.e., positionally consistent and similar in diction, such that users can see emergent relationships.

One prior piece of HCI work, ParaLib [51], does explicitly exploit these theories for system feature design, but does this in the domain of code.

There are two hypothesized benefits of this view. One is based on an understanding of human perception: the grid layout should help users compare more LLM responses because the spatial arrangement assists their memory. The other benefit is based on Variation Theory, which posits that discerning the impact of a critical aspect, for example model temperature, is only possible when experiencing variation along that dimension, isolated from variation along other dimensions.

Given that the features implemented in this work are in line with design implications of Variation Theory and Analogical Learning Theory, the results suggest that there may be further utility of these theories for guiding the design of future systems that help users make sense of data and form mental models from examples.

Theories of human concept learning suggest that a key step in forming accurate, robust mental models of a phenomenon is to be able to discern the underlying dimensions of variation (Variation Theory) and any latent structures beneath superficial details (Analogical Learning Theory). By detecting and communicating which sentences are both structurally analogous (by virtue of their position within the response) and semantically related (by virtue of highly overlapping content), users should be able to more easily identify emergent structures, as well as compare and contrast particular compositions of structural elements across responses and syntactic elements that may vary in meaningful ways across analogous sentences within those responses. These theories assert that these subtasks are key ingredients in forming those robust accurate mental models, i.e., learning from the LLM responses in order to better perform their overarching task.

In this work, in line with Variation Theory, the existing and novel features instantiated and described in the next subsection collectively identify patterns of consistency, variation, or both; they are explicitly designed to make emergent dimensions of consistency and variation easier for the user to perceive.

Variation Theory describes how helping people perceive the different dimensions of consistency and variation across examples (here, LLM responses) of the object of learning helps them more quickly and robustly leap to more accurate mental models. Analogical Learning Theory describes how people can form mental models or schema from perceiving structural analogical relationships across superficially varying examples (again, here LLM responses).

Variation Theory [35] and Analogical Learning Theory [13, 14] each propose mechanisms for how people may conceive and update their mental models based on concrete examples, or use their mental model in new situations.