The concept of “Social Intelligence” seems to come up a bunch, both in the context of autism and in the context of a bunch of socioeconomic research. But I think existing social intelligence scales are kind of bad. To understand how, let’s contrast them to the notion of general intelligence.
IQ tests try to measure the g factor, cognitive abilities that are useful across a wide variety of tasks. To do so, they can just grab a bunch of cognitive tasks and average the performance across them together, and this tends to give a reasonable measurement of the g factor, precisely because it is so general.
One might want to do the same with social intelligence, grabbing a bunch of social tasks and evaluating performance over them. But one can’t really do this because social tasks are “big”; they involve interacting with other people, but it is expensive to have another person around, and it takes a long time to build enough relationships with them to perform any sort of meaningful task.
My view of existing social intelligence tests is that they address this by postulating some “core capabilities” which are used in social tasks, and then they aim to measure those core capabilities rather than social performance. For instance, “Mind In The Eyes test” assumes that a core capability of social intelligence is being able to read someone’s mind from the expression they give with their eyes.
At its core, there are basically two kinds of social intelligence tasks, differing in the ways they measure their “core capabilities”. “Trait-based” social intelligence tasks measure them in basically the same way that personality tests do, and so they basically end up being personality tests that have been relabeled to be “social intelligence”. Meanwhile, “ability-based” social intelligence tests measure them in basically the same way that IQ tests do, and so they basically end up being IQ tests that have been relabeled to be “social intelligence”.
Ultimately, because social tasks are “big”, I think a social intelligence measure must be “formative” like the above. However, without a fairly accurate objective outcome criterion, it is hard to know what exactly to measure (what is redundant, important, missing, etc.). Lately I’ve been thinking that it must be possible to do way better with respect to outcome criteria. I feel like it should surely be possible to measure social performance by considering relationship history. That is, people may have relationships with their family, friends, coworkers, romantic partners and others, and it seems like it should be easy to assess how well those relationships are going, and use this as a ground truth criterion for a sort of “general social functioning”. And then it seems like it should be possible to study why some people do better or worse in relationships, and come up with a more specific measure of social intelligence which centers on the factors relevant to that.
Constructing a measure of social intelligence seems relevant to autism, as autism is often conceptualized as being closely related to poor social skills. It might also be relevant to personality disorders, as poor social skills might also contribute to them? Idk. I guess if this was to be done specifically for autism, one should probably first check to what extent autistic social relationship performance is worse than allistic social relationship performance. The effect size might be too small to be informative.
Also needs to account for any manifestation of the “double empathy problem”—if us autistic folk have some degree of ‘social intelligence’ that works perfectly well with autistic folk but falters with allistic folk, and vice versa, then what are we measuring?
An example might be, one allistic social intelligence test is to determine emotional state from the expression of the eyes, and … … here I realize that there’s not exactly a standardized way to correctly determine recognition of states like inanimate object feelings, and not everyone is lexythmic enough to score perception of their emotional state overall …
… well, it needs some workshopping. But given the potential extent it’s just tricky for minds that don’t think alike to connect socially, we want to be explicit about what we’re measuring; if that’s social relationship performance independent of allistic/autistic state, specifically our ability to perform social relationships with allistic folk, the difference, or what.
I think there’s two ways this would show up in the research process.
First, autistic people’s social outcomes would have the potential to become better than what one would naively expect, as their relationships with other autists would drag up their social outcomes scores. This would lead to such factors being weighted less when one studies factors which influence social outcomes.
Secondly, once some concrete social abilities are discovered, one can directly investigate whether they exhibit double empathy dynamics.
Oh also, another thing:
The concept of “Social Intelligence” seems to come up a bunch, both in the context of autism and in the context of a bunch of socioeconomic research. But I think existing social intelligence scales are kind of bad. To understand how, let’s contrast them to the notion of general intelligence.
IQ tests try to measure the g factor, cognitive abilities that are useful across a wide variety of tasks. To do so, they can just grab a bunch of cognitive tasks and average the performance across them together, and this tends to give a reasonable measurement of the g factor, precisely because it is so general.
One might want to do the same with social intelligence, grabbing a bunch of social tasks and evaluating performance over them. But one can’t really do this because social tasks are “big”; they involve interacting with other people, but it is expensive to have another person around, and it takes a long time to build enough relationships with them to perform any sort of meaningful task.
My view of existing social intelligence tests is that they address this by postulating some “core capabilities” which are used in social tasks, and then they aim to measure those core capabilities rather than social performance. For instance, “Mind In The Eyes test” assumes that a core capability of social intelligence is being able to read someone’s mind from the expression they give with their eyes.
At its core, there are basically two kinds of social intelligence tasks, differing in the ways they measure their “core capabilities”. “Trait-based” social intelligence tasks measure them in basically the same way that personality tests do, and so they basically end up being personality tests that have been relabeled to be “social intelligence”. Meanwhile, “ability-based” social intelligence tests measure them in basically the same way that IQ tests do, and so they basically end up being IQ tests that have been relabeled to be “social intelligence”.
Ultimately, because social tasks are “big”, I think a social intelligence measure must be “formative” like the above. However, without a fairly accurate objective outcome criterion, it is hard to know what exactly to measure (what is redundant, important, missing, etc.). Lately I’ve been thinking that it must be possible to do way better with respect to outcome criteria. I feel like it should surely be possible to measure social performance by considering relationship history. That is, people may have relationships with their family, friends, coworkers, romantic partners and others, and it seems like it should be easy to assess how well those relationships are going, and use this as a ground truth criterion for a sort of “general social functioning”. And then it seems like it should be possible to study why some people do better or worse in relationships, and come up with a more specific measure of social intelligence which centers on the factors relevant to that.
Constructing a measure of social intelligence seems relevant to autism, as autism is often conceptualized as being closely related to poor social skills. It might also be relevant to personality disorders, as poor social skills might also contribute to them? Idk. I guess if this was to be done specifically for autism, one should probably first check to what extent autistic social relationship performance is worse than allistic social relationship performance. The effect size might be too small to be informative.
Also needs to account for any manifestation of the “double empathy problem”—if us autistic folk have some degree of ‘social intelligence’ that works perfectly well with autistic folk but falters with allistic folk, and vice versa, then what are we measuring?
An example might be, one allistic social intelligence test is to determine emotional state from the expression of the eyes, and …
… here I realize that there’s not exactly a standardized way to correctly determine recognition of states like inanimate object feelings, and not everyone is lexythmic enough to score perception of their emotional state overall …
… well, it needs some workshopping. But given the potential extent it’s just tricky for minds that don’t think alike to connect socially, we want to be explicit about what we’re measuring; if that’s social relationship performance independent of allistic/autistic state, specifically our ability to perform social relationships with allistic folk, the difference, or what.
I think there’s two ways this would show up in the research process.
First, autistic people’s social outcomes would have the potential to become better than what one would naively expect, as their relationships with other autists would drag up their social outcomes scores. This would lead to such factors being weighted less when one studies factors which influence social outcomes.
Secondly, once some concrete social abilities are discovered, one can directly investigate whether they exhibit double empathy dynamics.