Biomechanics of Reading using wearable sensors

We will capture the reading process in a non-intrusive manner by collecting quantitative and qualitative data using wearables to gain further insight into human activity recognition.

Two-thirds of students in their first year of college are below the literacy levels by NAEP standards for GED proficiency. Furthermore, prose, quantitative, and document-based reading skills are significantly lower on average among students in two-year colleges than those in four-year colleges. The technology in our solution attempts to address the lack of research on reading habits with the presence of wearable sensors, aiming to make the user more present and engaged. This research would be able to assist in developing solutions for students struggling with literacy skills, perhaps by providing an increased understanding of the neuromotor requirements of reading. Our team is attempting to observe and systemize the collection of data on human behavior and activity recognition of reading in natural environments using the non-invasive and privacy-protecting advantages of wearable motion sensors on the wrists and books. This current field of research utilizes neuroimaging or direct observation as the main forms of analysis but they are not focused on the analysis of the natural processes of reading. These tools can alter an individual's behavior in ways that would not occur naturally in their absence. It is impossible to measure awareness of the reading process at our current stage, however, it is our desired outcome for future work to gain insight into reading comprehension.

Our primary solution is the utility of our database (not yet created) for individual readers to compare what their reading looks like in natural environments using non-intrusive technology. We are organizing pilot-level small group sessions at our community college to collect numeric and descriptive experimental data on reading habits using wearable motion sensors on each wrist and the spine of the book. During each trial, sensors will collect directional and inertial motion data from the three points of reference, creating a biomechanical “picture” of the participant’s reading session. The sensors will use a gyroscope, accelerometer, ambient light, and temperature for trial data during each reading session to be used for analysis with different machine-learning tools. With such information, we aim to provide more awareness of participants' personal reading habits and utilize this study and its technologies to gain insight into human behavior through activity recognition.

Our solution serves a diversity of populations. The pilot-level sessions will target students at our community college by inviting them to participate in research that may impact their experience with reading and generate new knowledge. Such information can address the needs of students who are underserved by providing educators insight into individual reading habits for possible appropriate intervention. This solution will also address their educational needs by bridging different sub-disciplines and research methods together to add to ongoing research in developing tools for reading comprehension, reading biomechanics, motor kinetics, machine learning tools, human activity recognition, and human behavior. Our solution also aims to provide a non-invasive tool that can collect and analyze information surrounding the biomechanics of reading habits for researchers, educators, students, and readers, and even invite new forms of curiosity to the conversation.

Our team is well-positioned to deliver this solution because we have started collecting initial data by running practice trials, using our research team members as the study participants. This valuable stage in our process is allowing us to test our technology, synthesize the most direct approach in using it towards our goals, and create instructional resources for future study participants. This research initially began through Alexandra’s work experience tutoring in the Norwalk Community College English Summer Bridge program which inspired the evolution of our college book club, now in its second year that continues to give students the opportunity to engage in reading and conversation outside of the classroom. Alexandra continued as a peer tutor for English and mathematics at our community college, and as a product of this experience, and of her own as a student of non-traditional age returning to the academic setting, she became interested in studying comprehension within the context of shame. Our team’s approach of putting the students’ personal volition at the forefront and maintaining a natural, non-invasive data collection method is consistent with that initial interest. Julia is our lead in design thinking, social entrepreneurship, and social media marketing. Additionally, she has been completing the suggested MIT Solve class for the marketing aspects of our solution. My interest in neuroscience and medicine prompted me to take part in this opportunity to ask difficult and possibly unanswerable questions about human behavior. We are also in the process of completing an Institutional Review Board Application through our community college to begin pilot-level trials with student, teacher, and staff participants. All of our team members have been certified to execute such research by "Protecting Human Research Participants Inc." which satisfies the National Institutes of Health requirements to perform studies involving humans.

The first step in our research was the simple observation of an unmet need within our institution. The average community college experience lacks a space for students to choose to read books without being awarded college credit and learn how to negotiate difficult conversations about those books without a moderator. In the 2021 fall semester, our team started a school-wide student book club at Norwalk Community College. We sought to address the need for a peer perspective within reading education at the school by building off the Common Read, a yearly institutional endeavor, in which students in First Year Experience courses read a book and participate in activities tied to its related themes. In this way, the book club utilized the embedded framework of the Common Read program to promote our initiative and gain support. By working within a preexisting framework towards achievable goals, we can turn our student-run book club into a platform for helping other students at different stages in academic development increase their reading comprehension skills. Through our experience in the book club, we have found that there is no equivalent in college classes for the skills that are developed in these student-run settings. Through this experience, we realized a further necessity for empirical data in order to move this solution forward. In keeping with our desire for a non-intrusive approach, we developed the idea of using biomechanical data to capture the reading experience. Our next step within this stage of our research is to use our reading model on the data collected and, bridge our research effort and the specific goal of this project: helping students. The book club at Norwalk Community College continues its success in its second year. Our team has been working with the MBient sensor system to identify its most effective placement sites for student use, such as the Tutoring Center and Counseling Center. Looking toward the next steps, the same concept is scalable at the other 11 community colleges in our Connecticut community college system; the organization of student book clubs is the primary initiative, followed by the introduction of the technology – both community use of the database and individual use of the wearable sensors. Expanding the reach of our population will increase and diversify our data collection, which will assist in our analysis and the generalizability of the results.  

What makes our solution innovative is the non-intrusive approach experimental method which sets us apart from current research on this question. In addition to setting up an experimental method that mimics natural reading conditions, our solution anticipates the use of innovative machine-learning techniques to analyze the biomechanical data collected. We recognize that if all machine learning tools return to the mean by regression, we will have to innovate to “see” those similarities and differences in this naturalistic reading experience. Our experimental design implements our technological model in a way that would not significantly disrupt users' behavior compared to traditional forms of observation. While machine learning models are often trained to remove anomalous data, treating all data points that do not adhere to the line of regression as “noise,” we plan to work within this so-called “noise,” using the complex representation of the natural reading process to the advantage of our analysis. Through this approach, we will be able to answer our initial questions surrounding the mechanisms of reading with more accurately reflecting data. A triangulation model was created to provide a data analysis framework that will be used to identify unexpected observations within outcomes. An additional innovative aspect of this model is the ability to still collect and logically analyze data for basic interpretation data that can be then compared with those substantial efforts within the somatic sensory and somatic motor research neurophysiology field.

In the long run, we believe our work has the potential to solve problems of current interest and lasting consequence, so that when submitted for peer review, our model and technology can then be implemented as a novel reading assessment and with possible utility in the education technology market. Our team's impact goals for the next year are to complete trials with participants, complete logical data analysis, and aim to implement our work on a larger scale at our community college and possibly others in our state. 

The team's solution is primarily based on two innovative applications of existing technology and additional supporting technology.

1. Our triangulation model capturing the biomechanics of reading involves the use of MBient MetaMotionS wearable sensors attached to straps on each of the participants' wrists and one on the book's upper spine. We will be collecting data on a directional and inertial motion from three points of reference with each sensor, creating a biomechanical “picture” of the participant’s reading session, as well as ambient light and temperature data for descriptive analysis.

2. Our success at the analysis level will require that we wrestle to acquire and identify unique signatures within the biomechanics of reading. This means that we must investigate novel machine-learning applications for intricate data analysis so that we can possibly find the “magic in the noise.”

3:  Additional Supporting Technology: The role of books as ancestral technology is crucial in our solution in investigating the biomechanics of reading and human activity recognition. We will also be utilizing tablets for connecting sensors, logging data, and sharing data through CSV files. 

As a proof-of-principle, the participants currently executing our solution for the first round of trials are the five members of the research group. In the next year, 20-50 participants will be recruited from our community college population including classrooms, student groups, and clubs.

Anticipating IRB approval, there are no current funding opportunities at our community college for additional resources and that is why we are submitting this application. We do foresee some difficulty with the analysis of the data, but this will not inhibit the transition between proof of principle, to pilot, larger analysis groups, and eventually, with more support the database of “biomechanical behavior while reading using a three-point geometric model” will then support the objective description of what reading looks like in our experimental context. We are not aware of any legal, cultural, or market barriers and we will seek outside counsel at such times when and if required.

N/A

N/A

N/A