By Deborah Levy
Aphasia is an acquired disorder of communication—not of intellect—caused by damage to areas of the brain that support language processing (National Aphasia Association, 2024). Aphasia most commonly occurs after left-hemisphere stroke and is highly prevalent—nearly 2 million individuals in the United States alone are currently living with aphasia (National Institutes on Deafness and Communication Disorders, 2025). While people with aphasia often participate in research studies, it is exceedingly rare for that research to be accompanied with any aid to the article that could be easily read by an individual with aphasia (think approximately 5 in 21,000 papers per a recent PubMed Search; Kasdan et al., 2025)—that is, the research remains largely inaccessible to precisely the people it is about.
Aphasia is an acquired disorder of communication—not of intellect—caused by damage to areas of the brain that support language processing (National Aphasia Association, 2024). Aphasia most commonly occurs after left-hemisphere stroke and is highly prevalent—nearly 2 million individuals in the United States alone are currently living with aphasia (National Institutes on Deafness and Communication Disorders, 2025). While people with aphasia often participate in research studies, it is exceedingly rare for that research to be accompanied with any aid to the article that could be easily read by an individual with aphasia (think approximately 5 in 21,000 papers per a recent PubMed Search; Kasdan et al., 2025)—that is, the research remains largely inaccessible to precisely the people it is about.
Though there is research on how to make complex written materials more aphasia-friendly, the process of creating accessible research materials can be difficult and time consuming to complete and is not yet a common practice, despite researchers wanting to engage meaningfully with people with aphasia and their care partners (Kasdan et al., 2025). To respond to this issue, and with the support of the AI Lab Seed Grant, my co-investigators and I are developing an AI-powered tool—Article Friend—to help researchers studying aphasia to efficiently convert their articles into aphasia-friendly, accessible versions.
Traditional scientific version
Aphasia-friendly version
Levy, Kasdan, et al., 2022. Perspect ASHA Spec. Interest Groups.
Traditional scientific version
Aphasia-friendly version
Individuals with post-stroke aphasia tend to recover their language to some extent; however, it remains challenging to reliably predict the nature and extent of recovery that will occur in the long term. The aim of this study was to quantitatively predict language outcomes in the first year of recovery from aphasia across multiple domains of language and at multiple timepoints post-stroke. We recruited 217 patients with aphasia following acute left hemisphere ischaemic or haemorrhagic stroke and evaluated their speech and language function using the Quick Aphasia Battery acutely and then acquired longitudinal follow-up data at up to three time-points post-stroke: 1 month (n = 102), 3 months (n = 98) and 1 year (n = 74). We used support vector regression to predict languageoutcomes at each timepoint using acute clinical imaging data, demographic variables and initial aphasia severity as input. We found that ~60% of the variance in long-term (1 year) aphasia severity could be predicted using these models, with detailed information about lesion location importantly contributing to these predictions. Predictions at the 1- and 3-month timepoints were somewhat less accurate based on lesion location alone, but reached comparable accuracy to predictions at the 1-year timepoint when initial aphasia severity was included in the models. Specific subdomains of language besides overall severity were predicted with varying but often similar degrees of accuracy. Our findings demonstrate the feasibility of using support vector regression models with leave-one-out cross-validation to make personalized predictions about long-term recovery from aphasia and provide a valuable neuroanatomical baseline upon which to build future models incorporating information beyond neuroanatomical and demographic predictors.
Levy, Kasdan, et al., 2022 Brain Communications.
Currently, Article Friend uses the Large Language Model (LLM) GPT-4o (along with the APIs for Google Docs and The Noun Project to generate accessible versions of scientific abstracts that follow best practices in generating aphasia-friendly materials, for example using simplified language, bulleted take-home points, bolded key words, and supportive visual icons (Herbert, 2012). When the prototype version of this tool was presented at this year’s Aphasia Access Leadership Summit in Pittsburgh, Pa. — a biannual gathering of speech-language pathologists, researchers, and individuals with aphasia— it was met with gasps of excitement and multiple rounds of applause. (See video below.) This warm reception was echoed in the comments and emails we have received since, and suggests to us that this truly is a project worth further pursuing!
While the future of Article Friend is looking bright, there is still work to do to develop the prototype into its culminating product. Some of the technical challenges that we’re currently facing include dealing with mismatches of visual icons with written words, oversimplifications of scientific content, and finite capacity for handling longer inputs. For these reasons, Article Friend is currently limited to processing scientific abstracts only, instead of complete research articles. These are obstacles we’re actively working to overcome.
Finally, we should note that, even when this tool is functioning at its best, we remain thoughtful about the fact that even the most sophisticated AI tools are still extremely prone to hallucination and—to borrow a Princetonian’s term—”bullshit” (Frankfurt, 1986; Hicks et al., 2024; Metze & Weise, 2025). It is crucial for our purposes—namely, ensuring that individuals with aphasia get access to not just accessible, but high-quality, information on research related to their condition—that for the foreseeable future this tool remains targeted at individuals with research expertise, who can effectively check its output and tweak phrasing and icons appropriately. Our goal with this tool is not to remove researchers from the loop, but simply to provide them with an easy way to get started making their work accessible to the people it affects the most.
Citations:
- National Aphasia Association: What is Aphasia?
- National Institute on Deafness and Other Communication Disorders: Aphasia
- A Practical Guide to Translating Scientific Publications Into Aphasia-Friendly Summaries
- Stroke Association: Accessible Information Guidelines
- Springer Nature Link: ChatGPT is bullshit
- The New York Times: A.I. Is Getting More Powerful, but Its Hallucinations Are Getting Worse
Leave a Reply