Introducing People with ASD to Crowd Work
Kotaro Hara, Jeffrey P. Bigham · 2017 · Proceedings of the 19th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '17) · doi:10.1145/3132525.3132544
Summary
This paper explores whether crowdsourcing platforms like Amazon Mechanical Turk (AMT) can provide viable employment opportunities for adults with Autism Spectrum Disorder (ASD). With an estimated 50-75% unemployment rate among adults with ASD in the U.S., the authors investigate whether the remote, asynchronous, and non-verbal nature of crowd work might bypass some of the social and environmental barriers that make traditional employment challenging for this population. The researchers conducted a six-week iterative user-centered design study with three participants with ASD, all of whom had comorbid conditions including intellectual disability and ADHD. The first four weeks focused on understanding how participants navigated the AMT platform and performed various types of micro-tasks, including image classification, image description, image transcription, surveys, and writing tasks. The researchers systematically evaluated which task types were most accessible, measuring both completion time and accuracy across different task categories. Based on these findings, they designed and evaluated a prototype tool called Assistive Task Queue (ATQ) during the final weeks. ATQ was specifically built to address the challenges participants faced with task search and task complexity, automating the process of finding appropriate tasks and breaking them into simpler, discrete steps.
Key findings
All three participants could successfully complete crowd work tasks, though their abilities varied significantly depending on the task type and their individual profiles. Tasks requiring lower executive function — such as image classification, image description, and image transcription — were more accessible than tasks demanding working memory and text synthesis, like writing. Task simplification had a statistically significant positive effect: breaking image transcription tasks into smaller components improved both speed and accuracy. The simplified "mini" interface reduced task completion time significantly (p<0.001) and improved transcription accuracy (p<0.01). The ATQ prototype further improved performance, with all participants completing transcription tasks faster in the ATQ condition compared to the standard AMT interface (p<0.05). Task search was a major barrier, consuming nearly 10% of participants' working time on average, with one participant spending 14.6% of session time searching for tasks. ATQ addressed this by automatically crawling, finding, and accepting appropriate tasks. Participants remained motivated to pursue crowd work throughout the study, citing factors like earning money, doing interesting work, and learning new skills.
Relevance
This research is significant for accessibility practitioners because it demonstrates that crowdsourcing platforms can be adapted to serve as employment pathways for people with ASD — a population facing severe unemployment. The study provides concrete design principles for making micro-task platforms more accessible: simplify tasks into discrete steps, automate task discovery, and match workers to tasks suited to their abilities. These principles extend beyond crowdsourcing to any digital work environment. The findings also highlight the importance of considering cognitive load and executive function demands when designing accessible interfaces. However, the study is limited by its small sample size (three participants, all with comorbid conditions) and the controlled lab setting. The low hourly wages typical of crowd work (median $1.38/hour on AMT at the time) raise questions about whether this represents meaningful economic opportunity without significant productivity tools.
Tags: autism spectrum disorder · crowdsourcing · employment · micro-tasks · assistive technology · user-centered design · cognitive accessibility · workplace accessibility