Social Skills and Autism - a Double Challenge
Social Interactions are a challenge for everyone on the autism spectrum. Even the most talented often discover some of the basics only later in life. Temple Grandin, in a lecture I watched on Youtube, said that "It wasn't until I was in my 50's that I realized there were all of these secret face signals". In an interesting interview on WrongPlanet, Bram Cohen, the brilliant CEO of a multi-million dollar software company, describes a simlar epiphany about the existence of social clues on the face when he was 19 or so.
The Second Challenge - Making an Impact on the Playground.
Helping students improve their social interactions is also a major challenge. While many brilliant teams have proposed and tested many promising approaches and interventions, a key hurdle remains whether these efforts translate into improvements in the student's everday social interactions. This challenge is sometimes called "generalization", or more specifically, generalization to everyday life.
Early in my exploration, an articulate mom told me how after a long search, she finally found a great program, Michelle Garcia Winner's "Social Thinking". Her son, a very bright 10yo w/ Aspergers, understood intellectually what he was learning in the program, but she said, "it all goes out the window on the playground". This really struck me. Making some impact on the playground became a driving idea early in my design of FaceSay, before I even read about "generalization".
In Behavioral Analysis, generalization is where something learned and mastered in setting A with conditions 1,2,3, is then executed successfully in setting B and conditions 4,5,6. This is often examined in terms of what happens when the reinforcers and prompts are withdrawn, and what is the best way to schedule or vary the reinforcers, and what types of reinforcers lead to the best result in a fully naturalistic setting, with no artificial reinforcers, i.e. in everyday life. Some ABA researchers believe, for example, that while a Discrete Trial Training approach leads to the fastest acquisition/mastery of a skill in a training setting, it results in less generalization than the slower "Incidental Teaching" approach.
More broadly, I think of generalization as when someone internalizes something they learned and can (and will) use it in novel situations. It's important to measure behavioral outcomes, but my design goal is closer to an insight, an "aha" or an "I get it" moment, similar to Temple Gradin's and Bram Cohen's above and to an 11yo FaceSay study participant. A year after the 2007 study, she was interviewed and asked "What did you learn from playing FaceSay?". She replied, "I learned I need to look at both halves of the face. I've been looking at just the bottom half, at the mouth."
In Vitro Performance Gains do not Correlate with In Vivo Benefits.
After reading many papers and talking and corresponding with dozens of researchers, teachers, practioners and families, it seems clear that while mastery of social skills in vitro, so to speak, is regularly accomplished, in all but a few studies, there is no measured gain in everyday life. Some studies reported anecdotal evidence of generalization, or some in vitro generalization (i.e. generalization to another in vitro setting), but very few Randomized Controlled Studies have shown generalization to everyday life.
Studies of TeachTown, Transporters and MindReading, for example, show evidence of in-game generalization, e.g. to "unfamiliar animated faces", but conclude with calls for future studies to assess generalization to everyday life. A 2001 study by Bernard-Opitz, et al, evaluating a software intervention for social problem solving, is often cited as an encouraging indicator of the potential of computerized interventions; yet it asks at the conclusion "whether behavior learned in the computer setting generalizes to the real setting...".
These results - which I think of as "intellectualization without generalization" - seem to be common across a variety of interventions and schools of thought, from behavioral, cognitive behavioral, neurocognitive, etc. and across many types of interventions, from social skills groups, parent driven, technology interventions and in the field, in schools, in the labs and in homes.
Here's a great quote (my emphasis added), from a leading team of Yale researchers in the opening paragraph of their "Enactive Mind" paper (see sidebar for a variety of other quotes)
it's a general missing skill set of being able to read faces, and being able to express thoughts on one's own face properly (there are other social cues, of course, but in my case I learned starting with the face, and everything else was easy from there)
A neuroscientist, in reference to an eye gazing study with Adults with Aspergers, stated that it was almost too easy to improve social skills in a training environment, but that the drop-off was dramatic once a real person was introduced. "They didn't just go from looking a lot to looking less. With someone else in the room they did not look at them at all."
"One of the issues that I have with his teachers at school is that they feel he understands... which he does if you are discussing situations in a clinical type environment. Transferring that 'understanding', to a real incident for him is a different matter.- Parent, on tigerfan.com
"Key issues discussed...included the need to focus on social skills in particiular ... emphasizing the need to employ effective intervention packages that stress generalization skills..."
"Overall, the [social skills groups] programs resulted in little change in the targeted behaviors, and students did not apply the skills outside the programs" Professor Bellini comments: "These results
"One of the most intriguing puzzles posed by individuals with autism is the great discrepancy between what they can do on explicit tasks of social reasoning (when all of the elements of a problem are verbally given to them), and what they fail to do in more naturalistic situations (when they need to spontaneously apply their social reasoning to meet the moment-by-moment demands of their daily life (Klin et al. 2000). While even the most intellectually gifted individuals display deficits in some complex social reasoning roblems (Happe 1994; Baron-Cohen et al. 1997), some, particularly those without cognitive deficits, can solve such problems at relatively high levels (Bowler 1992; Dahlgren & Trillingsgaard 1996) without showing commensurate levels of social adaptation. This discrepancy is troublesome because, while it is possible to teach them better social reasoning skills, such new abilities may have little impact on their real-life social or communicative competence (Ozonoff & Miller 1995; Hadwin et al. 1997)."
FaceSay improved social interactions on the playground in an RCT
In a 2007 Randomized Controlled Study (N=49), in the blinded observations of interactions with their peers on their normal school playground (ala Hauck, Feinstein, et al. 1995), the FaceSay group initiated more social interactions with their peers, made more eye contact, and exhibited fewer negative behaviors. This was a suprising and by far the most significant result.
In addition, the FaceSay group also improved on standard Ekman emotion recognition measures (an in-vitro measure). This is rather unexpected, given that FaceSay helps students modulate and discriminate between facial expressions, but never uses any labels. "Happy", "Sad", etc. are never mentioned. They also improved on in-vitro face recognition tests (Benton). The higher functioning students did not, but the lower functioning students did improve on social validity measures, the parent self reports.
The control group in the study used Tuxe Paint, a fun drawing PC program, so they also got pulled out of the classrom (always motivating :-) ) and got some extra attention.
I am sometimes asked why or how FaceSay helps generalization to the playground. The truth is I have my hunches (and pending patents :-) ), but we don't know. Success in the medical field sometimes comes from a "cocktail" approach, yes? Well, FaceSay is a "stew" I synthesized from dozens of ideas from Ekman's Action Units, Video Modelling, Enhanced Perceptual Functioning, Central Coherence, Behavioral Analysis, Cognitive Behavioral Therapy, Embodied Cognition, writings from autistic adults, Montessori, Human Computer Interaction, many innovative projects and papers I stumbled upon (thank you google :-) ) and my own imagination and observations.
A few studies are on the drawing board to try to identify some of the contributing elements, from studies of reinforcers, to studies that I hope will identify the least expensive "active ingredients" that will deliver a benefit. I am focusing more on advancing what else can be done. I recently released the option for video-realistic personalized avatars. I think this will further increase the benefits in everyday social interactions. And I have dozens of innovative ideas pending.
"although Type III measures [e.g. cognitive social skills measures] have some face validity, they do not predict important social outcomes for students."
"Further, there is evidence that skills may be displayed in laboratory/clinic settings, but not necessarily applied in the child’s daily life at school or home. Generalization and flexible skill use in natural environments continues to be a challenge"
"We saw some anecdotal evidence of generalization [to everyday life], but the study data did not rise to the level of significance."