Examining Entropy and Complexity of Digital Texts

• In Event: A Multimodal Mindset: Examining 21st Century Educational Problems Through a Critical Digital Literacies Lens

Abstract

Purpose of the Study
The widely-adopted Common Core explicitly emphasizes the quantitative evaluation of text complexity, providing a list of titles that illustrate such (corestandards.org). However, it does not include a single example of multimodal literature, which leaves the impression that reading is exclusively a language-based activity. This may be due to the fact that there are no standard metrics to analyze multimodal readability (Serafini et al., 2018). Transinformation analysis offers a solution.

Theoretical Framework
Transinformation analysis (Weltner, 1973) is rooted in information theory (Shannon & Weaver, 1949). Weltner (1973) proposed it as a method to measure readability, which he defined as “the interaction between a reader and a given text” (p. 120). Measured in units of bits per seconds—entropy—it aims to find the difference between the objective information in a message and the subjective information received by the recipient.

There is an established relationship between entropy and sentence comprehension (Hale, 2006; Shannon, 1951). Furthermore, research on entropy and salience indicates that low complexity images require less cognitive processing to comprehend (Kadir & Brady, 2001). Therefore, the entropy values yielded from transinformation analysis may be a strong indicator of readability.

Research Methodology
This study used previously collected data on multimodal reading (Author, 2012). The participants comprised 15 eighth-graders who earned “Advanced” scores in reading on a state standardized test. Data documents consisted of think aloud screen recordings of the readers interacting with Episode 2 of Inanimate Alice and transcripts from post-oral retellings that were conducted after the think aloud sessions. Think aloud data and mouse tracking data were collected to determine the readers’ areas of visual attention. Transinformation analysis was performed at the pixel unit level to measure the entropy of the document and attentional areas of the readers.

Findings
Findings showed that Inanimate Alice is a complex text and multimodal reading a complex task. The strong traditional readers were weak multimodal readers, and the reasons for this appear to be multi-layered. The above average complexity of the document likely presented a high cognitive load, exacerbated by the split-attention effect layout on a number of the screens. High cognitive load may have motivated the readers to selectively attend to lower entropy information (i.e., linguistic mode). The predominant tendency to attend largely to the linguistic mode may have been further influenced by perceptions that reading is a language-based activity. As such, this study yielded rich insight into multimodal reading behavior, and transinformation analysis proved a viable method for analyzing multimodal readability.

Significance

With today’s focus on STEM education, these findings are alarming because multimodal reading is foundational to learning in STEM subjects (Costa, 2020). A continued absence of multimodal documents in the classroom confounds the problem, which may be due to the perception that fewer words makes multimodal documents less complex. Transinformation analysis disproved that misperception, demonstrating that they are in fact quite complex; and though this study was experimental, it proved the strength of transinformation analysis as a method, as a tool, and as a contribution to theory.

Author

• Amanda Hovious, Emporia State University