Background
I have been collecting data on a test where I ask people to rate their social relationships. Each social relationship is rated on a scale from 1-9, across five facets of "helping with tasks", "keeping contact", "sharing knowledge", "sharing resources" and "sharing opportunities and connections", with people being informed that 1 means no function at all and 9 means like a hivemind.
I have asked people to rate four different types of social relationships across the above criteria: family, friends, coworkers and romantic partners. To ensure comparability of scores, I filtered people for having a job and a romantic partner. There are various ways to obtain "overall scores" from this data, but for the purposes of this market I will just do a simple average.
I feel like logically speaking of the concept of "social intelligence" or "social skills" is "real", then it should have a big effect on average social performance. So long-term, I hope that this average score can be treated as some sort of "social general intelligence factor" (at least if one finds some way of adjusting for environmental factors).
Autistic people are typically considered worse at social stuff than allistic people, so to investigate the extent to which this is reflected in social performance, I am about to recruit a sample of autistic people for my study. I tried to balance the sample so about half were diagnosed in childhood and about half were diagnosed in adulthood. I also balanced the sample by gender.
Resolution criterion
I will compute the social performance scores for the allistic sample and for the autistic sample, and then compute Cohen's d (that is, group difference / within-group standard deviation), with negative numbers indicating that autistic people perform worse socially, and positive numbers indicating that autistic people perform better socially.
If the Cohen's d is in the [-3, 3] range, this market resolves to that Cohen's d. If the Cohen's d is less than -3, it resolves to -3. If the Cohen's d is greater than 3, it resolves to 3.
If it is discovered that there is some bias that leads to misestimating the autistic people's social performance (e.g. high support needs autistic people are maybe not answering internet surveys? e.g. autistic people might be marginalized and so have lower expectations for social relationships, making my crude social functioning measure overestimate their functioning?), this market will resolve to the misestimate, rather than attempting to adjust for it.
1,000
3.00Final results!
On a scale from 1-9, allistic people rated their relationships at 6.11, and autistic people rated their relationships at 5.95, a -0.15 group difference, which with a standard deviation of 1.08 means a Cohen's d of -0.14.
Cohen d by relationship type:
Family: +0.14
Friends: -0.21
Coworkers: -0.42
Partner: +0.25
Cohen d by interaction type:
Helping each other: -0.11
Maintain relationship: -0.33
Share knowledge: -0.32
Share resources: -0.05
Share opportunities and connections: 0.13
@tailcalled I find it surprising/interesting that autistic people performed so well. Suggests to me that I don't understand autism (again despite being autistic) and/or don't understand social performance. Probably both.
Also I don't really recommend that people read too much into these results. The sample sizes are low so my main takeaway is "there's no huge difference between autistic people and allistic people in quality of social functioning", and I don't really make anything of the specific pattern of results. (Though I think it was worthwhile to double-check by type. Like it could have been that there would be a huge difference by relationship type or interaction type. It's just that there wasn't.)
Your main effect would most likely be one of the following:
Survey response bias - you have no objective measures in your survey, and so if autistic responses are more realistic and allistic responses are more optimistic, you'd capture that effect instead of any actual differences in behavior or performance. Note that this is not necessarily related to a bias in the response rate. That's a different can of worms, and what would matter is whether people who chose not to respond were drastically different across the sample categories, not whether the autistic and allistic response rates differ.
Selection bias - if your sample includes only those who are employed and in a relationship, you are choosing people who are not in the lowest bracket for functioning on either of those core areas. If instead you surveyed the partners, that might improve the validity slightly, and likewise if you included factors such as "time in a relationship" or "relative performance at work" (measuring the social quality of their workplace may be difficult, but promotions and leadership roles tend to imply non-terrible social perception by coworkers, and it's worth noting that plenty of people would prefer not to have social relationships with their colleagues).
@TamarSpoerri I'm expecting that the results will probably be fairly close, but it's not my strongest position, just a feeling really, and it's a cool market so I wanted to join. I'm more than happy to lose mana on this market to more knowledgeable predictors!
@TamarSpoerri based on the autistic people I have engaged with in my life, I'm guessing that
"helping with tasks", will be similar
"keeping contact", will be negative
"sharing knowledge", will be positive
"sharing resources", will be similar
"sharing opportunities and connections", will be slightly negative
The feeling is just based on the interactions I've had, it's very biased, I wouldn't put too much confidence in it, and I would have never bet if I didn't see it linked in discord and think it was a cool market that deserved more engagement.
@higherLEVELING I have data on 45 allistic people, and am aiming to get 18 autistic people, but might have to settle for 16 if the remaining 2 responses don't come in.