doi.org/10.1177/13623613251325934
Canfer Akbulut & Geoff Bird
An analysis of more than 5 million tweets created a huge data set that could be investigated. It allowed a statistical test to be run to see the similarities between what autistic people write, what their parents write, what is written by autistic self-advocacy groups, and what is written by autism advocacy groups that autistic people do not lead.
It found two things:
- Autistic self-advocacy (ASA) groups and autistic self-advocates show closer semantic alignment – that is how meaning can match up. Words shared between ASA groups and autistic people appeared in more similar linguistic contexts than those same words shared between ASA groups and parents of autistic children.
This suggests that autistic self-advocacy organisations’ language tracks more closely with that of autistic people themselves.
- Non-self-advocacy (nASA) organisations and parents of autistic children show closer semantic alignment. In a parallel result, the words used by nASA organisations were more similar to those used by parents than to those used by autistic self-advocates.
This indicates that these non autistic led groups may be more attuned to non-autistic family members’ perspectives.
Statistically, all these effects were robust: none of the true differences emerged in random shuffles of the data. In other words, the patterns are not just flukes but indicative of genuine differences in whose language is represented more similarly by which organisations.
If – and there are two big ifs – Twitter is representative of community views, and large language classifications don’t introduce error – then this shows that different types of advocacy organisations map more closely onto different segments of the community. Autistic-led groups align more with autistic people, while non-autistic-led organisations map more closely to the language patterns of parents.
The results of this study don’t clearly show whether there is a failure in how organisations represent different groups, because many important details are unknown. We don’t know how decisions were made, or if an approach was taken to widely define the community. To find this out, the next step may be to look at organisations’ representational claims and bring in insights from groups about what representation they expect to see.