Disability Services (DS) are the most common strategy for ensuring that students with disabilities have access to higher education. These offices provide accommodations for students that can apply to both physical (e.g., building) and digital (e.g., course content) accessibility. Prior work has shown that the access needs of disabled students are not always successfully provided, and if they are, they are not always accomplished in a timely manner. Consequently, there may be dissonance between when professors believe an accommodation has been fulfilled (e.g., when a PDF they posted has been made accessible by DS) and when it is actually completed. While prior work has documented the accessibility challenges faced on university campuses and with disability services, work has not yet focused on how to better support DS, student, and faculty interactions and communication. This study aims to do so by investigating how communication currently flows, the issues that arise within interactions and implementations, and whether technical solutions can better support interaction for the accommodation fulfillment process.

Social robots have been used to support mental health, because they are capable of forming meaningful interactions with humans due to their social intelligence, which boost user engagement with social robots. Thus, social robots can better foster emotional support with mental health data visualization compared to other technologies in collecting and displaying relevant data from humans. In this Human-Robot Interaction (HRI) research work, I explored their potential as community-based tools rendering data visualizations that represent a community’s mood patterns. Collecting and visualizing the mood patterns of a community might help the community members raise awareness about the mental health of the community, and whether there are significant factors negatively impacting the mental health of community members. In a two-day study, twelve participants recruited from a university community engaged with a robot displaying mood data. I created a data visualization template to show random mood data and placed it in a social robot. The design choices (e.g., spectrum of data varieties, data representation) of the template were made hypothetically in a random manner with no evidence basis from any particular community’s moods. During the study, participants were asked questions verbally and through written prompts on paper to express their feelings about robot-rendered visualizations’ usefulness in their lives. Given the feedback from participants, I modified the data visualization on the robot screen with the aim of increasing accessibility, universality, and usefulness of such visualizations. Currently I am implementing a real-time data collection and visualization system in the robot’s end-user-programming tool. In the near future, I plan on conducting studies to assess how interacting with the robot’s automated data collection and visualization software affects users and if it produces an improved experience for them. Eventually, I hope to evaluate how data visualization enhances a sense of belonging and support among community members.

This work proposes a novel generative design tool for passive grippers—robot end effectors that have no additional actuation and instead leverage the existing degrees of freedom in a robotic arm to perform grasping tasks. Passive grippers offer interesting trade-offs between cost and capabilities. However, existing designs are limited in the types of shapes that can be grasped. This work proposes to use rapid-manufacturing and design optimization to expand the space of shapes that can be passively grasped. Our novel generative design algorithm takes in an object and its orientation with respect to a robotic arm and generates a 3D printable passive gripper that can stably pick the object up. To achieve this, we address the key challenge by jointly optimizing the gripper shape and the insert trajectory to enlarge the set of objects that can be pasively grasped. We evaluate our method on a testing suite of 23 objects, all of which were evaluated with physical experiments to bridge the virtual-to-real gap. Inspired by the true cost of repurposing infrastructures in assembly lines following the recent changes in demand early in the COVID-19 crisis, our work allows a cost effective solution to rapidly generate, fabricate, and deploy custom passive grippers on the existing robot arms for the new products in demand.

Heart failure (HF) is a leading cause of hospitalizations in the United States, with nearly 900,000 hospitalizations annually and more than 500,000 new diagnoses each year. Many HF patients suffer from elevated venous pressure due to a failure to adjust medications when necessary, causing nearly 1 in 4 HF patients to be readmitted within 30 days of discharge. However, to monitor venous pressure, physicians currently use Central Venous Pressure (CVP), which must be measured in-clinic. We propose a low-cost, remote method of estimating CVP in-home utilizing a standard smartphone camera with the goal of preventing unnecessary readmissions. By taking daily, noninvasive measurements of CVP, patients may be provided with an early indication to adjust medications to maintain safe venous pressure levels, both improving their quality of life and lowering the cost of care. To measure CVP using a smartphone camera, our research team is developing a mobile application to capture video recordings of a subject’s neck. These recordings are then analyzed using our signal processing pipeline, which aims to filter for venular pulsation resulting from changes in pressure due to respiration. We expect to find varying amplitudes of light captured at different elevations of the body, and based on these amplitudes, we theorize that we can determine levels of CVP. These CVP estimations will be correlated with CVP measurements taken using current standard practices by physicians. We anticipate that in a future study, HF patients may utilize this monitoring tool to compare in-clinic CVP measurements during hospitalizations and follow-up visits to our estimated CVP measurements.

The NIH 3D Print Exchange is a public and open source repository for primarily 3D printable medical device designs with contributions from expert-amateur makers, engineers from industry and academia, and clinicians. In response to the COVID-19 pandemic, a collection was formed to foster creative submissions of low-cost, locally manufacturable personal protective equipment (PPE). To understand trends from this extraordinary occurrence of medical making, we performed a mixed-methods analysis of this collection. We used a combination of qualitative data from a thematic analysis and quantitative data from web scraped details of over 600 submissions. From this analysis, we found a disconnect between the NIH’s intention for the platform and how it was used. Instead of generating a diverse array of designs, the submission requirements and rating designations led to a rapid convergence of the design space. In this presentation, I present our findings for what we believe resulted in this disconnect and provide suggestions for how to improve upon the repository’s design. This work contributes valuable insights into the outcomes of distributed, community-based medical making and how platforms can support regulated maker activities in high-risk domains such as healthcare. Furthermore, many of our recommendations could be applied to non-health focused maker repositories such as Thingiverse and Instructables.

Open-source communities are central to facilitating the development of collaborative computing projects online. In any community, there are often policies that govern how the community functions, and as such communities become larger, projects have slowly seen more variety in their governance systems, making it no longer possible to just assume the governance structure of a project. The emergent practice of GOVERNANCE.md files in online collaborative code repositories such as GitHub works to close that gap by clearly defining the governing rules of what role contributors play and what guidelines they should be following. In this project, we developed a python script to scrape all the public GOVERNANCE.md files on GitHub in order to understand what the landscape of governance on GitHub looks like according to these files. Then, we analyze these files by creating data visualizations to compare different features of these files and characteristics of the community or repository (number of collaborators, creation date, number of edits to file, etc). This can then help develop answers to other key questions—how community governance files and models have changed over time, what aspects have changed and the driving factors/rationale behind these trends, and whether or not the changes in these models are following any specific trends (such as becoming more democratized).

Audio description (AD), an additional narration track that conveys essential visual information in a media work, is imperative for improving video accessibility for people who identify as blind or low vision (BLV). The AD creation process includes three steps: writing the script, recording the voiceover, and mixing the narration track with existing video audio. Despite being the primary beneficiaries, BLV individuals are limited in how they can contribute to AD authoring by technology inaccessibility and societal biases. The blind community and sighted allies advocate for including blind individuals in the AD creation pipeline, as their expert perspectives lead to high quality descriptions that prioritize the needs of the BLV community. To contribute to providing equal access to AD creation and to combat stigmas against blind AD writers, I am (1) designing and prototyping semi-automated AD features to make authoring more accessible to BLV creatives and (2) assessing the efficacy of question and answer systems in providing context for independent AD script authoring by BLV writers. Existing literature regarding AD does not address the needs of blind writers and the ways in which they interact with audio description writing systems. As there are currently very few blind writers in the AD industry, I am testing this system with 2 blind writers and at least 8 BLV people who are interested in writing AD. My findings so far show that current AD writing systems are inaccessible, and that BLV writers seek to have a detailed objective understanding of characters, background settings, and cross-frame actions prior to crafting artistic descriptions. Through this project, I am extending knowledge on BLV engagement with audio description creation and offering design considerations for AD writing systems to be more accessible for BLV writers who wish to independently write audio descriptions.

Chronic pain is characterized by persistent pain following the healing process or even in the absence of injury. The neural basis of chronic pain has been linked to several structures and pathways in the brain. However, the underlying neural electrophysiology behind chronic pain is not yet well understood. Here, we present a novel pain intensity decoder that inputs segments of neural electrophysiology and outputs predicted pain intensity values. We work with epilepsy patients who have week-long electrode implants for clinical monitoring. Using historical patient-reported pain intensities, scaled from 0 to 10, we identified segments of neural electrophysiology time-synched to when patients reported their pain levels. From these time segments, we extracted neural features, such as power-in-band values from the delta (1-4Hz), theta (4-8Hz), alpha (8-12Hz), beta (13-30Hz), gamma (30-55Hz), and high gamma (65-115Hz) frequency bands. We then used these neural features as inputs to train supervised machine learning models to predict pain intensity. We expect that prediction accuracy will depend on electrode location, where increased accuracy would occur in pain-related regions such as the anterior cingulate and insular cortex. Our classification model demonstrates that pain can be successfully decoded from neural electrophysiology and indicates potential for informing pain interventions and treatment methods for chronic pain.

Misinformation is a major problem in today’s online information ecosystem. One way that platforms address this problem is by providing viewers with signals so they can assess the credibility of the content that they encounter. However, there are few existing signals on online video-sharing platforms, which is concerning as video platforms like YouTube are a common way for people to get information. We seek to understand how citations can be applied as credibility signals to videos on the video-sharing platform YouTube. First, in an interview study, we investigated how users interacted with existing credibility signals on YouTube, and found that video intent and context shaped how users would apply both existing signals as well as future citations. We used these findings to inform our design and implementation of a browser plug-in that allows users to create and view citations on all videos they see on YouTube. We now want to explore how users might use this system to apply citations on YouTube and how viewing citations on a video might impact user behavior. In this second study, we plan to observe how the system affects the user's perception of video credibility and the effectiveness of our system as a credibility signal. 

Side-channel attacks on typing input devices can leak private user data such as passwords or other personal information. We have developed a system to perform such an attack by exploiting the timing of keystrokes as an input signal. Prior work has shown it is possible to retrieve this timing information by exploiting weaknesses in operating systems that leak the timing information of hardware interrupts and cache accesses. However, others have demonstrated systems to counter this attack by injecting fake data at various points in the software stack. Other prior work has demonstrated it is possible to infer individual typed keys by analyzing the frequency domain of the sound emitted by each key press, but it has not yet been shown that this type of attack is possible at natural typing speeds. Additionally, homophonic keyboards would be resistant to an attack that exclusively uses acoustic frequency information. We propose an algorithm that uses a combination of the above approaches—infer the timing information of the key presses from an audio recording of a user’s natural typing, analyze the frequency content of the sounds, and predict the typed words. Our preliminary study shows that we can identify boundaries between words with 79% accuracy. We use machine learning models to predict the current word using its inter-keystroke timings and the predictions of previous words. Our initial study demonstrates 76% prediction accuracy on a fixed dataset of 30 words. Overall, we show that it is not sufficient to prevent leaking keystroke information within the device’s own operating system. We present a solely audio-based algorithm to identify typed words, so the ubiquity of other passive recording devices like smartphones and smart speakers may still allow us to identify a nearby user’s typing on a computer keyboard, an important security risk.

For over 40 years, various biological activities, like gene regulation, have been tracked by protein reporter systems. However, the number of uniquely addressable protein reporters that can be used together is limited due to their overlapping readout signal. This prevents simultaneous measurement of multiple reporters (multiplexing), which not only impacts scalability and convenience, but also the potential complexity of the system of interest. To overcome this problem, we have previously designed a new class of reporter proteins, Nanopore-addressable protein Tags Engineered as Reporters. These are more scalable and multiplexable than traditional reporter strategies and are read out on a commercial nanopore array. They are detected and distinguished based on their peptide barcode regions which yield distinct ionic current blockades as they dwell within the nanopore sensitive region. Here, we optimize this system for mammalian cell systems, investigate phosphorylation motif-barcodes, and analyze new random and designed barcode sequences. We demonstrate this system in human embryonic kidney 293 cells transfected with differentially barcoded genetic circuits. We aim to utilize this novel reporter system to investigate complex mammalian processes such as chromatin regulation dynamics and the determination of cellular phenotypes using customized reporter cassettes. Ultimately, our technology will increase the scale and complexity at which systems such as these can be studied, leading to deeper understanding of biological programming and thus more robust synthetic gene circuit development.

With the exponential growth of data that is used and stored in our world, new forms of data storage will be required as the capabilities of our existing electronics plateau. Synthetic DNA is an attractive medium for data storage due to its density and longevity. With the growth of DNA data storage technology, there comes the need for exploring DNA computation. The potential to form boolean circuits and neural networks has already been actualized using DNA hybridization and strand displacement circuitry. However, in order to use DNA based neural networks to solve larger computational problems, we must improve current circuit designs to become more robust and scalable. In prior works, all DNA species are in a single test tube; the avoidance of cross-interaction from different circuit components is achieved by designing orthogonal DNA sequences. However, because the number of orthogonal DNA species is limited, the size of the DNA neural network is constrained to be quite small. In order to solve these issues, we modularized the circuit using magnetic beads. Modularizing the design not only makes the circuit simple and robust, but it allows for automation using droplet-based microfluidic technology. In this work, we implement a modularized (droplet-based) digital Binarized Neural Network (BNN). We first confirmed that a single binarized neuron functioned as expected by measuring output fluorescence. Next, we implemented a single layer with 5-input neurons to confirm the functionality of the layer. Then, demonstrated the ability to cascade multiple layers by using the outputs of preceding layers as the inputs of subsequent layers. Finally, we automated the BNN on the Opentrons OT-2 robot. Using microfluidic technology we automated the experimental protocol, which is normally performed manually in a wetlab, and by doing so, we are one step closer to building a practical DNA-based computer system.