3.1. Visitors’ Interaction with the Microcosmos Table
Analysis of the month-long log sample, consisting of two million (2,070,350) log events, suggested that visitors interacted with all of the available cards, during 858 sessions. The average session time was 96 s, which compares favourably with other reported exhibit holding times in science center and museum contexts [11
]. Furthermore, visitors engaged all the available interactive features in the system throughout the study period, including all the biological theme categories. Finger movements covered an average accumulated distance of 4.6 m per session, and the majority of interactions with the table interface (approximately 56%) were performed using two fingers (e.g., zooming and rotating). This may indicate the intuitiveness of the zooming and rotation finger gestures for exploring the biological visualizations. Together with the large finger distance traversed on the table surface, this also suggests a potentially important role of bodily interaction in engaging and exploring visualized biological content for education [3
The session construct was applied since there are no “natural” units of analysis associated with the table or readily available in the data [26
]. This is a consequence of the multi-user and multi-touch features of the table, where usage may range from a single individual interacting with the table for a well-defined period to groups of individuals that interact with variable levels of cooperation (cf. [12
]). Also, the anonymous data does not contain any information about the composition of users at any one time. Figure 2
provides an example of the distribution of data points generated by users’ interaction with the touch table during one session. It is intended to demonstrate that the captured data relates to interactive finger-based gestures such as selecting, moving, zooming and rotating. In this example, multiple users (probably at least three, positioned at the top left, top right, and lower right corners of the table surface, respectively) are engaged with the exhibit during the session, and interacting across large parts of the touch table surface. The higher density of data points in the upper left corner indicates that a particularly active person (or persons) was positioned there, while a person(s) standing close to the lower right corner were less active.
In summary, results from the overall usage data analysis indicates that visitors engaged with the exhibit to an extent comparable to typical exhibits. In addition, they utilized the multi-touch functionality and actively interacted with the table.
3.2. Visitors’ Preferences for Visualized Biological Content
The log files contained an average of 5.53 category switches per session. In terms of biological content, the virus category was the most frequently activated (0.95 activations per session on average). The other categories were less frequently activated, with average number of activations per session ranging from 0.72 to 0.80 (Table 2
Interaction with individual cards varied greatly. For example, the number of sessions was seven times larger for the most frequently activated card than for the least frequently activated card. Similarly, the number of times a card was accessed first during a session ranged from 0 to 48, while interaction with cards yielded average numbers of log entries that varied by a factor of almost 5. The difference in average number of entries between the most (median = 141) and the least (median = 6) used cards was statistically significant (Mann-Whitney U = 551, p < 0.001). The differences in interaction between different cards indicate that there were systematic differences in how users chose to engage with the visualized content. In the following, trends in those differences will be described and interpreted.
The overall ranking score, based on the two attractiveness measures and the measure of holding power, yielded a list of the ten most highly accessed cards and the ten least accessed cards (Table 3
and Table 4
). Six of the most highly used cards (images 3, 4, 5, 7, 9, 10 in Table 3
) concerned viruses, while none of the least activated cards (Table 4
) concerned viruses. The only cards among the highly accessed cards that depicted cells did so in the context of virus infection with only parts of a cell visible (image 9 in Table 3
), while the three depictions of cells among the least used images and videos portrayed whole cell structures (images 55, 56, 60 in Table 4
). Apart from the difference in scientific content, there were also clear differences in representational style preference. Most notably, the highly accessed images and videos contain a larger number that integrate colour, sharper contrasts, and are brighter than the less accessed images and videos. For example, all four images among the highly accessed that are based on instrument output (e.g., electron microscopy) are colourful (images 1, 6, 9, 10 in Table 3
), while the three transmission electron microscopy images among the least accessed images have a greyscale appearance (images 56, 60, 61 in Table 4
). Interestingly, an exception to the trend towards lower brightness and colour intensity among less used cards are images that use a watercolour style to represent complex molecular scenes (images 57, 58, 59 in Table 4
A possible explanation for the revealed trends in engagement may be offered by construal level theory [27
]. According to construal level theory, the level at which humans construe objects and events is influenced by the psychological distance to that object, the distance from an egocentric reference point in the here and now. In particular, objects that are at a close psychological distance to the person will be construed at a low level, which means that concrete details and contextual factors will be emphasized in the construal. By contrast, objects that are at a larger psychological distance will be construed in a more abstract manner, emphasizing the central, overarching aspects of an object. Multiple psychological distances may be involved in this effect, including spatial, temporal, social, and hypotheticality distances [27
Lee et al. [28
] propose that black-and-white images may be associated with a larger psychological distance than colour images in at least two ways. Firstly, black-and-white may connote a greater temporal distance, given that historical events that took place before the invention of colour reproduction are often displayed in greyscale. Secondly, black-and-white may also connote a larger social distance, since most people experience their everyday surroundings in colour. In addition, they add that black-and-white may tend to highlight contours and shapes, which might reinforce the same cognitive processing that is associated with high-level construal [28
]. Taken together, this would tend to induce a high-level construal of black-and-white images and a low-level construal of colour images.
The above reasoning does not explain the appearance of brightly coloured images among the less engaged images. We propose that hypotheticality may be an important dimension in this regard. Objects that are psychologically close on the hypotheticality dimension are more real and associated with a higher probability, while imagined or unlikely objects and events are more distant [27
]. In relation to the present study, images that appear more like physical objects (e.g., 3D-models and electron micrograph images) may be perceived as more real (closer on the hypotheticality dimension) compared to 2D paintings (more distant on the hypotheticality dimension), which could induce a difference in construal level between different types of coloured images. The presence of one image that also uses a watercolour style to display a virus (image 4, Table 3
) indicates that other mechanisms may also be involved. For example, familiarity with the portrayed objects may influence the psychological distance in relation to another dimension, namely the information distance [29
]. It is reasonable to suggest that the spherical virus-particle portrayed in image 4 (Table 3
) is more familiar than the protein complexes shown in image 57, 58 and 59 (Table 4
), respectively. Pre-knowledge may also have other effects that are unrelated to the effects from psychological distance on construal level.
We argue that the findings in this study could be explained by differences in psychological distance associated with the images leading to different levels of construal. However, the ranking of images and videos was based on a measure of interactive engagement rather than level of construal per se. Hence, applying construal level theory to the results of the study would imply an association between level of construal and interactive behaviour. Given that a low-level construal tends to focus on details, it may be reasonable to expect that users would interactively explore the details to a higher degree than images that are construed at a high level. Conversely, a high level of construal focuses on central properties, for example form, and the relevant processing of the images may therefore not require as much interaction. Further research is needed to investigate the relations between levels of construal and interactive behaviour among users.