SUBSCRIBE NOW!

Vol. 40 No. 6 — Jul/Aug 2023

AI ETHICIST

AI FOR DIAGNOSIS AND LEARNING

Alas, many neurodivergent people lead lives of quiet desperation. They must navigate a world that is largely ignorant of, or indifferent to, their concerns and challenges. They are square pegs being shoved into round holes. This is an incalculable loss to society, but AI can be a powerful force for good in changing this status quo.

AI is beginning to help with diagnosis. Several AI tools—leveraging behavioral data, based on brain scans, or using speech patterns—are being developed. The FDA has approved a machine learning-based device for autism spectrum disorder diagnosis and another that diagnoses autism spectrum disorder based on eye tracking.

AI can also make education more effective. Social robots can help autistic children learn more easily, and analyzing teacher-student interactions in the classroom helps teachers craft their communication and enhance teaching effectiveness. AI for diagnosis and AI for learning are large areas of inquiry and not the focus of this article, but let’s note that AI is beginning to make an impact.

AI FOR  DIAGNOSIS AND LEARNING

“The Use of Artificial Intelligence in Screening and Diagnosis of Autism Spectrum Disorder: A Literature Review”
ncbi.nlm.nih.gov/pmc/articles/PMC7298904

“New AI-Driven Algorithm Can Detect Autism in Brain ‘Fingerprints’ ”
hai.stanford.edu/news/new-ai-driven-algorithm-can-detect-autism-brain-fingerprints

“AI Detects Autism Speech Patterns Across Different Languages”
news.northwestern.edu/stories/2022/06/ai-detects-autism-speech-patterns-across-different-languages

FDA Authorizes Marketing of Diagnostic Aid for Autism Spectrum Disorder
fda.gov/news-events/press-announcements/fda-authorizes-marketing-diagnostic-aid-autism-spectrum-disorder

“EarliTec Diagnostics Receives FDA Clearance for Autism Diagnosis Support Tool”
mobihealthnews.com/news/earlitec-diagnostics-receives-fda-clearance-autism-diagnosis-support-tool

“UNICEF Guidance on AI for Children: Application to the Design of a Social Robot for and With Autistic Children”
uwe-repository.worktribe.com/output/8808696/unicef-guidance-on-ai-for-children-application-to-the-design-of-a-social-robot-for-and-with-autistic-children

“An Artificial Intelligence Approach for Selecting Effective Teacher Communication Strategies in Autism Education”
nature.com/articles/s41539-021-00102-x

NEURODIVERSITY AT WORK

The estimate is that 15%–20% of the population is neurodivergent. It is safe to assume that we all have a friend or a family member who is neurodiverse. There is a strong case to be made for embracing neurodiversity in the workplace: Neurodivergent employees bring valuable, complementary, and unique talents to the table. Despite that, unemployment among neurodivergent people is very high; according to some estimates, in the U.S., up to 85% of high-functioning autistic adults are unemployed.

Practically all large organizations have diversity, equity, and inclusion (DEI) strategies. But it’s clear that there is a need for greater focus on neurodiversity, including DEI plans to plug the neurodiversity gap. AI can certainly bolster corporate neurodiversity. But it also requires anticipating and accommodating different sensory needs, communication patterns, visual processing, and social interaction habits in the workplace.

Recruitment

AI can help improve the recruitment process. Many recruitment practices unintentionally eliminate or are unfriendly to neurodiverse candidates. For example, job descriptions may emphasize skills extraneous to the role, such as a high degree of extroversion when the job does not require it. AI can help refine job descriptions to make them less exclusionary.

Traditional interviews assess not only the actual skills of the candidates, but also the intangible impressions: body language, eye contact, perceived confidence, etc. Such subjective inputs can result in swift elimination. AI-based online exercises can help provide a more objective assessment and level the playing field for neuro­divergent candidates.

Job Performance

AI can help improve job performance. As an augmented intelligence tool for neurodiverse employees, it can be a personal smart assistant—setting alerts, sending reminders, generating meeting summaries, helping compose emails, converting text-to-speech or speech-to-text, acting as a guide to surroundings, helping decode the communication patterns of colleagues, and more. All of these technologies exist already, and users just need to be trained on them. On that note, it’s just as important to train colleagues and managers on how to work best with their neurodivergent co-workers.

Companies are constantly reconfiguring business processes and work routines when they automate or outsource. Why not explore whether small changes to old work routines can help neurodivergent employees? Remote work arrangements can be an option in many cases—workplaces are certainly more used to hybrid work now compared to a few years ago.

NEURODIVERSITY  AT WORK

British Medical Bulletin: Table 2, Timeline of Neurominorities
ncbi.nlm.nih.gov/pmc/articles/PMC7732033/table/TB2

“Neurodiversity as a Competitive Advantage”
hbr.org/2017/05/neurodiversity-as-a-competitive-advantage

The Center for Neurodiversity & Employment Innovation
entrepreneurship.uconn.edu/neurodiversitycenter-2

NEURODIVERSITY AND AI ETHICS

AI can be a double-edged sword. Many current practices leave a lot to be desired because they don’t consider the perspectives of neurodivergent people. If that forms the baseline for AI tools, we are worse off. Adherence to ethical AI principles is paramount. AI creators may mean well, but they need more awareness and sensitization to the problems faced by the neurodiverse community and the alternative pathways they require—before the solutions are built. AI technologies must be co-created: Neurodiverse voices must be purposefully involved in their design and validation.

AI ethics must include neurodiversity as a priority. The field of AI ethics, rightly, emphasizes fairness, accountability, and transparency. The notion of fairness is about avoiding discrimination in automated decisions. The main lenses used to evaluate discrimination are race and gender. Discriminatory outcomes for people with disabilities, along with people who are neurodiverse, are not part of the mainstream AI ethics discourse. For example, disability-related keywords appear only 11 times in a total of 1,659 AI ethics articles reviewed by Assistive Technology in 2019. There are laws preventing discrimination on the basis of disability. The datasets used for evaluating whether AI systems are leading to discriminatory outcomes frequently lack neurodiversity data, and this needs to be remedied too.

Many perceived barriers are in the mind of the neurotypical people: that it takes too much effort and too many resources to attract, recruit, onboard, develop, and retain neuro­divergent talent. But making the workplace more neuro­diverse is a sound business strategy—legally required in many cases—and is our collective moral imperative. AI, when paired with human empathy, can get us there.

NEURODIVERSITY AND  AI ETHICS

“Automating Autism: Disability, Discourse, and Artificial Intelligence”
digitalcommons.odu.edu/sociotechnicalcritique/vol1/iss1/8

“Coverage of Ethics Within the Artificial Intelligence and Machine Learning Academic Literature: The Case of Disabled People”
pubmed.ncbi.nlm.nih.gov/30995161