Research Projects

We creating general accessibility solutions that empower people with disabilities to accomplish their goals, including programming, physical computing, and more. We also create novel fabrication technologies, including for accessibility. Our goal is to improve inclusion in and accessibility of our digital future. In addition, prior work done by our group focused on sustainability. A list of all publications in order of time can be found at the end.

Fabrication

accessibility | inclusion | all

President Obama once said ‘3D printing […] has the potential to revolutionize the way we make almost everything” [Obama 2013].’ One might say that computers have the same ability. Together, computational advances and fabrication can revolutionize personal problem solving, particularly with respect to accessibility. 

Our group is embarking on an effort to enhance fabrication technology, both through the use of new and better materials and in its ease of program. Much of our effort here is on embedding standard programming capabilities such as testability, modularity, iterability, and version control into the design and fabrication process.

Challenges at the intersection of fabrication and accessibility
A purple freestanding lace piece on top of a jar lid, adding a rigid, grippable base.

Shaping Lace

A computationally generated image of three different configurations of hangers and chains to support a bird feeder.

FabHacks with Everyday Objects

Ribbed and checkered knit colors generated programmatically using knitscript.

KnitScript: A Domain-Specific Scripting Language for Advanced Machine Knitting

Model of an automatically-generated flat splint with breathing holes, and the same splint after it has been printed, heated, and shaped around a thumb

Domain Specific Metaheuristic Optimization

A picture of two people wearing 3D printed PPE. On the left is a face shield (42.2% of submissions) and on the right a facemask (28.4% of submissions).

Rapid Convergence: The Outcomes of Making PPE during a Healthcare Crisis

Optimized tactile map

Maptimizer

More publications…

Accessibility

 fabrication | inclusion | all

One in 5 Americans has a disability, and the needs of people with disabilities are extremely heterogeneous. Yet accessibility is not a given, and assistive technology today is mostly homogenous and very expensive. Our group is focused on the scientific challenges that can address accessibility across a broad range of domains. A major focus of this work is creating general tools so that anyone can use tools such as 3D printing, knitting and other fabrication technologies.

“Policies and laws” are social determinants of health for access to telehealth care. “Policies and laws” intervene with the status quo and influence digital determinants of health for telehealth care; therefore, it can be considered a digital determinants of health of telehealth.

Telehealth and Digital Health Innovations

Partial of a table in the paper (listing the 4 epistemic positions) and partial view of the conclusion section of the paper.

What Counts as ‘Creative’ Work?

A table with an assortment of cane tips, cane handles and feet in various shapes.

“A Tool for Freedom”

Stylized illustration of a person with robotic features sitting at a café table, looking at a phone, with a glowing light around them. Four human figures sit at surrounding tables, engaged in conversation. The central figure has a robot icon on their chest and a speech bubble with the same icon, symbolizing artificial intelligence or feeling like a robot in a social setting.

Autoethnographic Insights from Neurodivergent GAI “Power Users”

Figure 1: The flowchart offers a map through the results. At the top we have communication access, which branches into two directions: 1) access needs (e.g., caption use, speechreading) and 2) language landscape (e.g., frequency, context, comfort, overall repertoire). There are two way arrows between these two, representing the question 'How do these interact/inform/shape each other?'. Below the question, as answer we have two groups: first, factors for a particular langauge (fluency, linguistic affordances, cultural perceptions) and second, practices across languages (translation, transliteration, translanguaging).

Toward Language Justice

Image showing four conceptual prototypes from the paper: participatory screenreader, interactive book and audio narrative, data jungle experience, @hackingaccess initiative

Cripping Data Visualizations

More publications…

Inclusion

accessibility | fabrication | all
Our group also works to address inclusion of other under-represented groups, including intersections of accessibility and race; discrimination and disability experiences of undergraduate students in the UW Experience Study and gender in academia and other settings.

Figure 1: The flowchart offers a map through the results. At the top we have communication access, which branches into two directions: 1) access needs (e.g., caption use, speechreading) and 2) language landscape (e.g., frequency, context, comfort, overall repertoire). There are two way arrows between these two, representing the question 'How do these interact/inform/shape each other?'. Below the question, as answer we have two groups: first, factors for a particular langauge (fluency, linguistic affordances, cultural perceptions) and second, practices across languages (translation, transliteration, translanguaging).

Toward Language Justice

Example messages about COVI-19 safety including "after we meet up, can you walk me to the bus stop? I've had some bad experiences, and as an asian-american woman I don't feel safe" and "I know everyone's vaccinated, but can we still be careful? A few things I had in mind..." and "These guidelines would help me feel safe: everyone is vaccinated; no physical contact; masks on; no sharing food/drink"

COVID-19 Risk Negotation

Artwork by Lydia X.Z. Brown,(they/them) 2020. Watercolor and ink drawing of four disabled people. The first is wearing a bipap respirator and a shirt that says disabled and proud! and has brown skin and dark brown hair in a bob. The second is wearing a green shirt that says Access is Love and has short green hair and green lips. They are in a wheelchair and have brown skin. Next is a person holding a white cain wearing a blue shirt saying "my other disability is a bad attitude" with shoulder-length brown hair and lighter skin. Finally, there is a person with dreadlocks in an orange shirt. Words like "disability" and "genius" are visible behind them. The bottom right of the image states "By Lydia X Brown, 2020. Honor crip, mad, disabled & neurodivergent wisdom & genius. Disability Justice Now! Downloaded from community.amplifier.org

Race, Disability and Accessibility Technology

Gender in Online Doctor Reviews

Passively-sensing Discrimination

Picture of potted plants and a bench with the word Africatown in the background, painted in bright red and green colors

Who Gets to Future?

More publications…

All recent publications

Return to top: accessibility|fabrication|inclusion|all

“Policies and laws” are social determinants of health for access to telehealth care. “Policies and laws” intervene with the status quo and influence digital determinants of health for telehealth care; therefore, it can be considered a digital determinants of health of telehealth.

Telehealth and Digital Health Innovations

Partial of a table in the paper (listing the 4 epistemic positions) and partial view of the conclusion section of the paper.

What Counts as ‘Creative’ Work?

A table with an assortment of cane tips, cane handles and feet in various shapes.

“A Tool for Freedom”

Stylized illustration of a person with robotic features sitting at a café table, looking at a phone, with a glowing light around them. Four human figures sit at surrounding tables, engaged in conversation. The central figure has a robot icon on their chest and a speech bubble with the same icon, symbolizing artificial intelligence or feeling like a robot in a social setting.

Autoethnographic Insights from Neurodivergent GAI “Power Users”

Figure 1: The flowchart offers a map through the results. At the top we have communication access, which branches into two directions: 1) access needs (e.g., caption use, speechreading) and 2) language landscape (e.g., frequency, context, comfort, overall repertoire). There are two way arrows between these two, representing the question 'How do these interact/inform/shape each other?'. Below the question, as answer we have two groups: first, factors for a particular langauge (fluency, linguistic affordances, cultural perceptions) and second, practices across languages (translation, transliteration, translanguaging).

Toward Language Justice

Image showing four conceptual prototypes from the paper: participatory screenreader, interactive book and audio narrative, data jungle experience, @hackingaccess initiative

Cripping Data Visualizations

A purple freestanding lace piece on top of a jar lid, adding a rigid, grippable base.

Shaping Lace

An AI generated decorative image of a person standing alone outside of a corporate boardroom, excluded.

Identifying and improving disability bias in GPT-based resume screening

This figure illustrates the process of our prototype system in translating an English sentence into an American Sign Language video. The top of the figure shows the example English text ("Do you commute to work by bike?''). Below it, the glosses and non-manual markers for each word in the sentence are displayed. The glosses are: IX-2p, TEND, COMMUTE, WORK, and BICYCLE, with a notation indicating a yes/no question. The next row contains a series of five diagrams representing five skeletal poses. Each diagram shows a specific handshape and location corresponding to the glosses and non-manual markers, color-coded for clarity. The bottom row contains images of a person performing the ASL signs, matching the glosses and non-manual markers above. In these images, the person is shown against a green background, demonstrating each sign with corresponding hand positions and facial expressions. Zoomed-in sections highlight the facial expressions associated with the yes/no question.

Towards AI-driven Sign Language Generation with Non-manual Markers

A computationally generated image of three different configurations of hangers and chains to support a bird feeder.

FabHacks with Everyday Objects

Wordcloud of alt text assigned to images in notebooks prominently featuring useless text such as "png" "image" "alt" and "Colab"

Notably Inaccessible

Left: A hand is shown moving back and forth on a laptop touchpad. Right: A TED video is playing and the slider is highlighted to indicate that the hands motion controls the video position.

Touchpad Mapper

A picture of a slide with the title "Animals that would be the best computer scientists" by G. Rade. To the right of the slide editor a "Slide Repair" panel is visible that shows a list of accommodations that can be applied such as changing the line spacing, font family, contrast, and colors

Bespoke Slides for Fluctuating Access Needs

Ribbed and checkered knit colors generated programmatically using knitscript.

KnitScript: A Domain-Specific Scripting Language for Advanced Machine Knitting

Model of an automatically-generated flat splint with breathing holes, and the same splint after it has been printed, heated, and shaped around a thumb

Domain Specific Metaheuristic Optimization

A lecturer showing a slide with an image of 3D printed gears on it. Students sit with their backs to us watching the lecture.

A Multi-StakeholderAnalysis of Accessibility in Higher Education

Example messages about COVI-19 safety including "after we meet up, can you walk me to the bus stop? I've had some bad experiences, and as an asian-american woman I don't feel safe" and "I know everyone's vaccinated, but can we still be careful? A few things I had in mind..." and "These guidelines would help me feel safe: everyone is vaccinated; no physical contact; masks on; no sharing food/drink"

COVID-19 Risk Negotation

Two AI-generated images show different interpretations of a doll-sized “crocheted lavender husky wearing ski goggles,” including one pictured outdoors and one against a wooden background.

Generative Artificial Intelligence’s Utility for Accessibility

14 drawings of participant gestures organized as 2 rows and 7 columns. The top row represents high agreement gestures, and the bottom row represents unique gestures that participants designed. The columns are functions (from left to right): move/pan, select, rotate, delete, open/close, zoom-in/-out, duplicate. For move/pan, the high agreement gesture was moving hand from left to right in mid-air and the unique gesture was flexing fingers from left to right. For select, the high agreement gesture was pointing with hand or finger, and the unique gesture was squeezing hand into fist. For rotate, the high agreement gesture was grabbing object and rotating about the person's wrist and the unique gesture was rotating shoulders. For delete, the high agreement gesture was swiping away and the unique gesture was looking down. For open/close, the high agreement gesture was swiping down, and the unique gesture was shrugging shoulders. For zoom-in/-out, the high agreement gesture was moving both hands towards and away from body, and the unique gesture was flexing shoulders in and out. For duplicate, the high agreement gesture was tapping twice in mid-air in two separate locations with the hand and the unique gesture was tapping chin twice.

How Do People with Limited Movement Personalize Upper-Body Gestures?

Artwork by Lydia X.Z. Brown,(they/them) 2020. Watercolor and ink drawing of four disabled people. The first is wearing a bipap respirator and a shirt that says disabled and proud! and has brown skin and dark brown hair in a bob. The second is wearing a green shirt that says Access is Love and has short green hair and green lips. They are in a wheelchair and have brown skin. Next is a person holding a white cain wearing a blue shirt saying "my other disability is a bad attitude" with shoulder-length brown hair and lighter skin. Finally, there is a person with dreadlocks in an orange shirt. Words like "disability" and "genius" are visible behind them. The bottom right of the image states "By Lydia X Brown, 2020. Honor crip, mad, disabled & neurodivergent wisdom & genius. Disability Justice Now! Downloaded from community.amplifier.org

Race, Disability and Accessibility Technology

A picture of part of a dashboard visualization showing sales by state, using a US map, bar charts, a pie chart, and total sales. To the side, part of a textual dashboard description is visible summarizing the same information visible in the charts. Above, a filter menu is visible where the user can select which region to show sales for.

Azimuth: Designing Accessible Dashboards for Screen Reader Users

A picture of a physical therapist helping a patient to use a stretch band

Physical Therapy Accessibility for People with Disabilities and/or Chronic Conditions

A picture of two mock heads showing how color and position can be used to visualize phonemes around the head. A green circle at bottom left shows the wh phoneme in the left image. A yellow blob at the top right shows the f phoneme.

The Role of Speechreading in Online d/DHH Communication Accessibility

GLOBEM logo, blue text with a world for the O.

Cross-Dataset Generalization for Human Behavior Modeling

PSST: Enabling Blind or Visually Impaired Developers to Author Sonifications of Streaming Sensor Data

A picture of two people wearing 3D printed PPE. On the left is a face shield (42.2% of submissions) and on the right a facemask (28.4% of submissions).

Rapid Convergence: The Outcomes of Making PPE during a Healthcare Crisis

A picture of a slightly see through shoulder and neck with ribs showing. A spaghetti strap is visible.

Chronically Under-Addressed: Considerations for HCI Accessibility Practice with Chronically III People

A complex diagram showing the flow of information from sensors through feature extraction, formation of a user behavior profile, and then final prediction.

Personalized behavior modeling: depression detection

A diagram showing how we build stress & resilience frameworks for student mental health. The framework has three components -- social characteristics, cumulative stress, and resilience resources. At the center where all three overlap is perceived stress, anxiety and depression.

Distress and resilience among marginalized undergraduates

A quote from the paper

COVID-19 and Remote Learning for Students with Disabilities

Optimized tactile map

Maptimizer

Knitted sweaters and associated patterns

Computational Design of Knit Templates

A barplot showing intervention acceptance for typout (57.2%) and typing only (48.8%) among 54 users over 10 weeks.

TypeOut: Just-in-Time Self-Affirmation for Reducing Phone Use

What Do We Mean by “Accessibility Research”?

Designing Tools for High-Quality Alt Text Authoring

Mixed Abilities and Varied Experiences in a Virtual Summer Internship

Anticipate and Adjust: Cultivating Access in Human-Centered Methods

Practices and Needs of Mobile Sensing Researchers

A lineplot showing anxiousness (Y axis, varying from 0 to 4) over time (X axis). Each student in the study is plotted as a different line over each day of the quarter. The plot overall looks very messy, but two things are clear; Every student has a very different trajectory from every other, with all of them going up and down multiple times. And the average, overall, shown is a fit line, is fairly low and slightly increasing (from about .75 to just under 1).

College during COVID

Sampler of PPE found on the NIH repository

Medical Making During COVID

(Left) Users can select accessibility qualities to filter the places shown on the map. (Right) Clicking on a place opens a popup window to see all of that place's coded accessibility qualities, pictures, reviews, and additional textual descriptions.

Navigating Illness, Finding Place

BLV Understanding of Visual Semantics

A user is tapping on their cheek. The tapping is sensed by a microphone in the earphone.

Interaction via Wireless Earbuds

HulaMove: Waist Interaction

Understanding Disabled Knitters

A picture of a knit lampshade in blue and green surrounding a lit lamp on an orange table.

KnitGIST: Generative Texture Design

Shows barchart of import of different features onetecting change in depression

Detecting Depression △

Gender in Online Doctor Reviews

A picture of Megan in her wedding dress leaning on her cane with a river and skyscrapers in the background. A second picture shows her partner working on the cane while wearing a mask and a faceshield for protection.

Living Disability Theory

Detecting Loneliness

The Limits of Expert Text Entry Speed

KnitPick: Manipulating Texture

Passively-sensing Discrimination

A venn diagram showing the domains of expertise of those we interviewed including people from hospitals, universities, non-profits, va networks, private practices, and government. We interviewed clinicians and facilitators in each of these domains and there was a great deal of overlap with participants falling into multiple categories. For example, one participant was in a VA network and in private practice, while another was at a university and also a non-profit.

Digital Fabrication in Medical Practice

A Makapo paddler in a one-person outrigger canoe (OC1) with the final version of CoOP attached.

Designing in the Public Square

Clench Interaction: Biting As Input

Picture of potted plants and a bench with the word Africatown in the background, painted in bright red and green colors

Who Gets to Future?

Splint that has been 3D printed in a material of an appropriate skin color and fit to a client's hand.

“Occupational Therapy is Making”

Sign showing: Sustainability -- Next Exit

Automatically Tracking and Executing Green Actions

Lyme Disease’s Heterogeneous Impact

Understanding gender equity in author order assignment

Picture of a 3D printed arm with backscatter sensing technology attached to it.

3D Printed Wireless Analytics

Interactiles

EDigs

Nonvisual Interaction Techniques at the Keyboard Surface

Picture of 3D models and a printout

Expressing and Reusing Design Intent in 3D Models

Picture of an elastic band with a sensor about the size of a belt buckle attached to it.

Hypertension recognition through overnight Heart Rate Variability sensing

The Tangible Desktop

Examples of 3d printed objects that are robust to measurement uncertainty.

Uncertainty in Measurement

Making the field of computing more inclusive for people with disabilities

Volunteer AT Fabricators

Tactile Interfaces to Appliances

Aversion to Typing Errors

Modeling & Generating Routines

The top of a young person's head with long hear, wearing a red felt crown with 3D printed decorations on it.

3D Printing with Embedded Textiles

Watch-ya-doin

Shows someone placing a pen in a cap with two different types of adaptations.

Printable Adaptations

A teddy bear wearing a knit hat, scarf (with pocket) and sweater

A Knitting Machine Compiler

Sketchup Interface with plugin for RapID being used to create a Tic Tac Toe board.

RapID — interactive RFID

Modeling Human Routines

Threadsteading

Diagram of FOCUS method for dynamic, cost-sensitive feature selection

Dynamic question ordering

3D printed attachments

Helping Hands

Supporting Navigation in the Wild for the Blind

I red bunny made of layers of felt fabric.

Layered Fabric Printing

Infant Oxygen Monitoring

3D Printed Prosthetics: Case Study

Images from a variety of projects supporting green behavior

StepGreen

Probabilistic Input

Severity of Chronic Lyme Disease

Replacing ‘Wave to Engage’ with ‘Intent to Interact’

Stepgreen.org

Automatically tracking green actions

A picture of the clavicle and neck of a person with translucent skin. The rib bones are visible through the skin.

Competing Online Viewpoints and Models of Chronic Illness

Access SIGCHI protest

Disability Studies and Accessible Technology Creation

Exiting the cleanroom: On ecological validity and ubiquitous computing