SLP Full Disclosure Ep. 117: Confronting Bias and Exploring Artificial Intelligence in Speech Therapy
Bias in speech and language screenings is a topic every speech-language pathologist (SLP) is familiar with but may feel hesitant to address openly. Admitting that bias exists in the tools and decisions we trust is both necessary and challenging. Once recognized, bias compels us to reimagine how we approach assessments, treatment, and patient interactions.
Through this blog, recapping episode 117 of the SLP Full Disclosure podcast, we’ll explore what bias looks like in speech-language pathology, discuss tools like normed tests and dialectal variations, and consider how artificial intelligence (AI) offers opportunities to confront and mitigate bias in clinical practice.
Whether you’re an SLP working with pediatric clients, a speech-language pathology assistant (SLPA), or a clinical fellow just beginning your career, this article will shed light on how we can use mindfulness, systematic improvement, and technology to ensure more equitable and accurate care.
Understanding Bias in Speech-Language Pathology
Before addressing bias in speech therapy, we must define what bias is. Bias refers to a systemic deviation from fairness or objectivity. It’s not inherently “bad”—bias is a natural human tendency, shaped by our backgrounds, experiences, and cultural norms. However, in the clinical context, unchecked biases can influence the outcomes of screenings, evaluations, and treatment plans.
Unintentional Bias in Action
For instance, many standardized tests assume that all children are familiar with specific vocabulary, objects, or cultural experiences. A child who hasn’t been exposed to certain words or objects might perform poorly not because of a language deficit but because of a lack of familiarity. Dr. Marisha Speights of Northwestern University highlights this in her research on dialect and norming samples, explaining that our tools often reflect limited definitions of what we consider "typical development."
The takeaway here is simple but powerful. Bias is inevitable, but identifying where and how it exists within assessments is the first step toward addressing it.
Expert Insights from Dr. Marisha Speights
Dr. Marisha Speights, a speech scientist and speech-language pathologist at Northwestern University and our special guest for episode 117, has dedicated her research to mitigating bias in speech assessment tools. She explains that many standard language screenings have historically been normed on predominantly white, English-speaking populations.
This limited sampling can lead to “outdated” or incomplete concepts of typical speech and language development. For example, older assessment tools might still include pictures of objects like rotary phones or typewriters, which current generations aren’t familiar with. Such seemingly small issues can unfairly penalize children from specific demographics and regions when these tools do not account for cultural and linguistic diversity.
Cultural Representation Matters
A critical question raised by Dr. Speights is this: Who are these tools designed for? And do they genuinely represent the linguistic and cultural diversity of the children we work with?
Normed tests often lack representation of bilingual children, regional dialects, and various socioeconomic contexts. Consequently, entire populations might be miscategorized as below-average language performers, leading to over- or under-diagnoses of speech and language disorders.
Addressing Dialect Variations in Assessments
Another significant source of bias stems from dialectical differences. Dr. Speights emphasizes that SLPs must account for dialect variations, such as African American English (AAE) or Southern American English, when interpreting test results. Dialects are not disorders but reflect rich linguistic diversity. Misinterpreting these variations as impairments leads to inequitable treatment.
Practical Steps for SLPs:
- Educate Yourself: Seek out training on cultural and linguistic differences relevant to your population.
- Critically Analyze Assessment Tools: Review manuals to understand the demographics represented in norming samples.
- Engage Parents and Communities: Gather feedback from caregivers about the child’s linguistic environment for a more holistic understanding.
SLPs who adapt their evaluations to consider these variables provide more accurate and meaningful diagnoses, ultimately ensuring better outcomes for the children they support.
The Role of Artificial Intelligence in Speech Therapy
AI is revolutionizing many industries, including healthcare. For speech therapy, this technology holds immense promise. Dr. Speights, who leads research in using AI-driven tools at Northwestern University, explains that artificial intelligence can augment traditional screening by identifying patterns clinicians may not notice.
How Artificial Intelligence Works in Speech Therapy
At its core, AI algorithms recognize patterns within the speech signal. These include subtle acoustic features that human ears might overlook. For example, AI can analyze biomarker data such as frequency, volume, or timing to provide detailed insights into speech development.
AI tools currently show a 96% accuracy rate in distinguishing between children with and without speech disorders. However, the challenge lies in ensuring that these tools are free from bias. Biases ingrained in AI systems mirror those inherent in the data used to train them. This makes it critical for researchers and clinicians to work together in designing equitable AI systems.
AI as a Tool to Mitigate Bias
When properly developed, AI has the potential to reduce inequities caused by human bias. For instance, AI could identify cultural and linguistic variations in speech without falsely categorizing them as disorders.
However, SLPs must remain critical and ask key questions about the AI tools they use, such as:
- Who developed this AI tool, and does the team include linguistic and cultural experts?
- Was the training data representative of the population I serve?
- What measures have been taken to address bias in the algorithm?
By staying informed and curious, clinicians can ensure that AI enhances, rather than hinders, equitable care.
A Balanced Approach to Technology and Testing
Bias in speech therapy is no longer a taboo topic. Through education, reflection, and advocacy, SLPs can make strides in minimizing its impact. AI, while not a perfect or standalone solution, offers opportunities to reduce systemic bias and improve the accuracy of our assessments.
To help guide this push for more thoughtful practices, follow these steps as an SLP or SLPA:
- Be Curious: Continue learning about how bias affects your tools and practices.
- Be Critical of Normed Tests: Review manuals, norming samples, and scoring guidelines.
- Leverage AI Thoughtfully: Use AI tools as a supplement to clinical expertise.
- Engage Communities: Seek input from caregivers and local communities to better understand the needs of those you serve.
The future of speech therapy calls for collaboration between clinicians, researchers, and AI developers to truly create inclusive and unbiased tools. For more insights into this topic and innovative solutions in speech therapy, tune in to Episode 117 and subscribe to SLP Full Disclosure today to stay informed about the most relevant trends and insights impacting the world of speech therapy.
