Difficulties with social interactions are common for autistic adults. They report few close friendships (Howlin et al., 2000; Orsmond et al., 2004) and are more likely to experience social exclusion and low quality of life compared to adults with cognitive or other developmental disabilities (Orsmond et al., 2013; DaWalt et al., 2019). These outcomes are even found for autistic adults without intellectual disability (Farley et al., 2009; Howlin and Moss, 2012; Lord et al., 2020), and appear largely independent of a person's autistic traits (Magiati et al., 2014). In fact, autistic people commonly experience similar, or even worsening, social disability as adults despite a measured reduction in autistic traits from childhood to adulthood (Howlin et al., 2013). Although previous work has primarily attributed interpersonal difficulties in autism to intrinsic deficits in social cognition and behavior (Baron-Cohen et al., 1985; Oberman et al., 2005), more recent empirical advances (Sasson et al., 2017; Morrison et al., 2019a; Crompton et al., 2020b), and informed expertise from autistic people (Yergeau, 2013; Milton and Sims, 2016; Gillespie-Lynch et al., 2017; Kapp, 2019; Raymaker et al., 2020), have increasingly highlighted the role of bi-directional factors, including inhospitable social environments, and the behaviors of non-autistic (NA) people, that also contribute to poor social experiences for autistic people (Milton et al., 2013; Sasson et al., 2017; Morrison et al., 2019a; Crompton et al., 2020b).

This reframing of interpersonal difficulties in autism from individual to relational is exemplified by The Double Empathy Problem (DEP; Milton, 2012). The DEP eschews traditional deficit-model explanations for the social difficulties autistic people often experience in favor of a transactional explanation, driven by a mutual breakdown of communication between people with different modes of social communication and understanding. In contrast with the decades of research documenting autistic difficulties inferring the mental states, emotions, and intentions of NA individuals (Baron-Cohen et al., 1985; Schultz, 2005; Morrison et al., 2019b), a growing empirical literature grounded in the DEP framework has found that NA adults make similar social cognitive errors when trying to understand their autistic peers (Edey et al., 2016; Sheppard et al., 2016). These misperceptions can lead NA adults to view autistic people unfavorably (Alkhaldi et al., 2019), and may contribute to social exclusion and poor mental health among autistic adults (Mitchell et al., 2021).

A bi-directional difference in communication styles for autistic and NA adults is further supported by differences in interaction outcomes for mixed vs. matched neurotype interactions. Within dyadic interactions consisting of either two autistic adults, two NA adults, or an autistic adult paired with a NA adult, both autistic and NA individuals showed a greater interest in future interactions with individuals who shared their neurotype (Morrison et al., 2020). Qualitative reports from autistic adults suggest that this preference may relate to an increased understanding and acceptance of autistic communication styles in interactions between autistic individuals (Crompton et al., 2020a). Indeed, a study of information transfer between autistic and NA adults (Crompton et al., 2020b) found that chains of alternating autistic and NA adults experienced greater communication difficulty than chains consisting entirely of autistic or NA individuals, which did not differ from each other. Collectively, these findings highlight the ways in which a mismatch between autistic and NA communication styles can impact autistic-NA interactions, and suggest that the “fault” of interaction difficulties between autistic and NA partners does not lie with either person alone, but in the intersection between the two.

Traditional deficit-model frameworks of autistic interaction difficulties have almost exclusively centered treatment on the autistic person via social skills and social cognitive training, with the implicit assumption that teaching more normative modes of social understanding and behavior will translate into improved social outcomes. These interventions have generally failed to produce lasting benefits for autistic adults (Bottema-Beutel et al., 2018), and may unintentionally encourage the masking of autistic ways of being (Pearson and Rose, 2021), increase internalized stigma (Botha and Frost, 2020), and contribute to depression (Cage et al., 2018), anxiety (Hull et al., 2021), and even suicidality (Cassidy et al., 2020) in adulthood. Furthermore, because many autistic individuals consider autism to be central to their identity (Botha et al., 2020; Crompton et al., 2020a), interventions designed to alter their core characteristics have been criticized as unnecessary or even abusive (Milton, 2014; Kirkham, 2017; McGill and Robinson, 2020).

Therefore, given that deficit-model treatments for social disability among autistic adults are minimally effective at improving life outcomes, and may in some cases harm mental well-being, alternative approaches for improving interpersonal difficulties between autistic and NA adults are beginning to be considered and tested (Jones et al., 2021). One potential avenue capitalizes on recent findings suggesting that improving autism knowledge and acceptance among NA individuals shows promise for reducing biases toward autistic children and adults (Gillespie-Lynch et al., 2015; Dickter et al., 2020a) and increasing inclusive attitudes (Jones et al., 2021). However, the benefits of autism training may not extend to all forms of bias (Dickter et al., 2020a; Bast et al., 2021; Jones et al., 2021), and it is unknown whether previously reported benefits translate beyond experimental settings to real-world interactions between autistic and NA individuals.

In our previous study (Jones et al., 2021), NA participants viewed a brief autism acceptance training (AAT) video, then rated their first impressions of videos of autistic adults, answered questionnaires assessing their autism stigma, perceptions of autistic abilities, and autism knowledge, and completed an implicit association test (IAT) to measure their implicit biases about autism. We found that compared to adults who completed a general mental-health focused training, as well as those in a no-training condition, participants who completed AAT had more positive perceptions of autistic abilities, greater interest in interacting with autistic individuals, and less autism stigma. However, implicit biases did not differ significantly across training conditions, suggesting that the training may have a limited impact on more subtle or covert forms of bias. The current study seeks to expand our previous work to a real-world setting to evaluate whether a brief AAT module for NA adults can lead to more positive interactions between autistic and NA adults. Non-autistic adults were assigned to either an AAT condition or a no-training control condition, with participants in the AAT condition initially viewing a 25-min video featuring factual information about autism and firsthand accounts from autistic adults (Jones et al., 2021). Non-autistic participants across both conditions were then paired with an unfamiliar autistic adult and completed a 5-min unstructured dyadic interaction, with participants blinded to their partner's diagnosis. Following the interaction, participants responded to questionnaires assessing their impressions of both the interaction and their interaction partner. Based on our previous findings (Jones et al., 2021), we predicted that NA participants in the AAT condition would rate their autistic partners more favorably and would have greater interest in interacting with them compared to NA participants in the control condition. At the level of the interaction, we predicted that NA participants in the AAT condition would feel closer to their autistic partners and would rate the interaction as higher quality compared to NA participants in the control condition. Because the training was designed to target the NA person, hypotheses centered on their responses to autistic partners, but the employed dyadic analyses also examined whether autistic adults evaluated NA adults who completed the training more favorably than those who did not. If supported, these hypotheses would provide evidence for the use of this training as a brief, accessible tool to improve interactions between autistic and NA adults.

Although previous research has demonstrated that training programs designed to increase autism acceptance and knowledge among NA people can reduce biases and improve inclusive attitudes toward autistic people (Gillespie-Lynch et al., 2015; Dickter et al., 2020a; Jones et al., 2021), no study to date has investigated whether training benefits extend to real-world interactions between autistic and NA people. The current study examines whether an AAT module previously shown to reduce autism stigma among NA adults and increase their interest in interacting with autistic adults presented in videos (Jones et al., 2021) produces improvements to interaction quality and partner evaluation during actual conversations between unfamiliar autistic and NA adults.

Compared to a no-training control condition, both autistic and NA adults reported greater social interest in one another following a “get to know you” conversation when the NA adult had completed AAT. Specifically, both autistic and NA participants in the AAT conditions expressed an increased desire to hang out with their partner in the future, suggesting that AAT not only improved NA adults' social interest in their autistic partners, but also increased their perceived social desirability among autistic participants. Thus, an acceptance training focused solely on NA participants produced a relational effect, leading to social improvements for both partners. Importantly, this improvement occurred despite participants' unawareness of their partner's diagnosis, suggesting that it was not influenced by demand characteristics. This result replicates a previously observed effect of AAT, in which the training increased NA adults' interest in hanging out with autistic people viewed in video clips (Jones et al., 2021), but extends it to real-world interactions with autistic people and, importantly, suggests it may also transfer to increased social interest among autistic adults in their NA partners. Although this result indicates that a brief and relatively easy-to-administer training for NA adults may increase mutual social interest among unfamiliar NA and autistic adults, it remains unclear whether the effect would produce sustained contact and relationship development beyond the experimental session. It is also unknown what, if any, aspects of NA behavior and communication differed following AAT and contributed to increases in social interest. Future work is encouraged to attempt to both replicate this effect and measure whether and how training alters NA behavior within interactions with autistic people. Research examining the impact of AAT on social interactions is limited, but additional training with a greater focus on autistic communication and expressivity may improve NA understanding of neurodivergent interaction styles. Because NA adults have been found to misinterpret autistic communication styles (Brewer et al., 2016; Edey et al., 2016), resulting in a breakdown in communication (Crompton et al., 2020b), how a double empathy focused training may affect perceptions of interaction quality for autistic and NA adults is worthy of further examination.

In contrast to this finding, no effects of training were found on the other three behavioral intention items. However, these items—living near, sitting near, or having a conversation with the person in the future—represent relatively superficial forms of social interaction that can occur with acquaintances or even strangers (Morgan, 2009), whereas the intention to “hang out with” and spend one's free time with another person reflects a closer level of contact associated with the development of friendships (Hays, 1989; Sias and Cahill, 1998), and may be a strong indicator of intimacy, particularly among males (Wood and Inman, 1993; Floyd, 1995; Floyd and Parks, 1995). Autistic individuals often have limited social opportunities (Lord et al., 2020), and can experience difficulties forming friendships (Mazurek, 2014), in part due to how they are perceived by others (Sasson et al., 2017). By increasing interest in future close interactions between autistic and NA adults, AAT may offer potential for improving social opportunities for autistic adults within NA environments.

Independent of training effects, several differences emerged between NA and autistic participants in their evaluations of each other and the interaction. For NA participants, positive ratings of their partner on many first impression and interaction items were associated with higher ratings of interaction quality. In contrast, interaction quality was largely unrelated to how autistic participants evaluated their NA partner. Similarly, NA participants but not autistic participants who perceived their partner to have greater intelligence in turn showed greater social interest in them. This may suggest a greater connection between person and interaction evaluation for NA compared to autistic adults. Such an interpretation is consistent with prior research showing stronger associations between trait evaluation and social interest among NA than autistic people (DeBrabander et al., 2019), and may indicate that trait judgments like awkwardness, likeability, and attractiveness are less relevant to autistic adults than NA adults when judging interaction quality. This interpretation—that autistic individuals place less weight on surface-level traits of their partner when evaluating interactions—is also supported by previous literature suggesting that shared interests rather than individual traits are more of a primary driver of successful friendships for autistic adults (Sosnowy et al., 2019). Future studies investigating other differences in interaction and friendship preferences between NA and autistic adults may highlight additional sources of relational disconnect that, through awareness and understanding, may offer avenues for improving interactions, inclusion, and social outcomes for autistic people.

This interpretation is further supported by the dissociation in ratings made about one another by autistic and NA partners within dyads. Only ratings of warmth significantly correlated between partners, and this correlation was negative, suggesting a disconnect between NA and autistic people in a fundamental aspect of interpersonal assessment. These results are notably different from those found in a previous study of real-world interaction among and between autistic and NA adults using the same outcome measures (Morrison et al., 2020), in which dyadic partner ratings for first impression items, behavioral intentions, interaction quality, and closeness were all significantly related. However, Morrison et al. (2020) included autistic-autistic dyads and NA-NA dyads, in addition to the mixed dyads used in the current study. It may be the case that, consistent with a DEP framework (Milton et al., 2013), inter-partner agreement increases in interactions between people of a shared neurotype and declines within cross-diagnostic interactions.

Contrary to prediction, AAT largely did not affect trait evaluations made of and by autistic adults, nor did it affect participant assessments of interaction quality. While null findings should be interpreted with caution, several factors may account for the lack of training effects for some ratings. First, although a larger sample size may have revealed more effects, it is also likely that a brief, one-time presentation may be insufficient for eliciting the large-scale behavioral changes needed to improve interpersonal perceptions within these interactions. Indeed, previous work suggests that the effects of AAT modules, including the one used in this study, may affect explicit but not implicit biases (Bast et al., 2021; Jones et al., 2021). Implicit biases are automatic, unconscious forms of bias that can contribute to unfavorable judgments about groups of people, and as such, the persistence of these biases previously shown to be prevalent among NA participants toward autism (Dickter et al., 2020b), may have impacted behaviors and attitudes toward autistic interaction partners. Importantly, not all effects of training were beneficial. One unexpected result was that autistic and NA partners in the training condition rated one another as less trustworthy than in the control condition. Perhaps AAT influenced NA behavior in ways that were unappealing to autistic participants and/or raised suspicions among NA adults about their interaction partners. Alternatively, this could be a spurious finding related to using distinct participants in the two training conditions or to the lack of a pre-test/post-test design. The employed analyses did not implement a correction for family-wise error, so future work should examine whether this effect replicates.

While participants' diagnostic status was not disclosed, and only two autistic participants chose to disclose their diagnosis, it remains possible that social desirability biases in NA participants may have influenced the results of the study. Both groups were exposed to the possibility of an autistic conversation partner, but the autism-specific training video may have primed participants in this condition to expect an autistic partner, leading to more favorable ratings in this condition compared to a no-video control condition or a non-autism related control. If present, social desirability biases may represent a potential strength, as the methodology of this study maps onto how similar trainings may be administered in the real world, with participants aware that the training is designed to improve their interactions with autistic people. Regardless, training effects were not consistent across conditions, including a potentially negative finding of reduced trust in training condition participants, suggesting that results were not driven solely by social desirability biases. Additional research examining if and how demand characteristics influence training outcomes are encouraged.

Additionally, NA attitudes toward autism were not assessed prior to participation, so these may have differed between the two training groups, minimizing the potential benefits of AAT. The young adult NA sample included in the study also may have already been more familiar with, and accepting of, autistic differences than the general population (White et al., 2019). This is particularly relevant for the current NA sample given that it was drawn from a university with one of the largest number of autistic students in the United States (Hoffman, 2016). Further, the diagnostic status of participants was also not disclosed to conversation partners. While this lack of disclosure can be viewed as a methodological strength, producing fewer demand characteristics and more ecologically valid interactions, effects may have been larger if disclosure had occurred, as has been found in previous studies (Sasson and Morrison, 2019). Because the training was compared to a no-training condition, as opposed to an active control, it is difficult to ascertain whether any effects were specific to the training video used here, or rather an effect of training in general. Comparison to a more generic training, such as that used in our previous study (Jones et al., 2021), may illuminate the unique benefit of autism-specific training. Perhaps most importantly, the sample in this study consisted exclusively of White, self-identified males to control for confounding effects of cross-gender and cross-race interactions. Given the impact of gender (Milner et al., 2019; Lai and Szatmari, 2020) and racial biases (Giwa Onaiwu, 2020; Jones et al., 2020) on the experiences of autistic adults, the results of this study may have differed in important ways for a more diverse sample. Finally, participants were young adults with verbal IQs in the average range, so the impact of AAT may not generalize to the broader population of autistic people, including non-speaking people, those with an intellectual disability, or older adults. Therefore, while the current study offers proof of concept for analyzing the effects of NA training on improving interactions for autistic people, future work should consider how NA perceptions of, and behavior toward, autistic people intersect with other salient aspects of identity not examined here.

These limitations notwithstanding, this study provides further insight into the relational aspects that contribute to interaction difficulties between autistic and NA adults and offers some limited support for the benefits of AAT for NA adults. Autism acceptance training in this study was associated with a greater future interest in hanging out for both autistic and NA adults within dyads, but the benefits of training did not extend to other ratings, including evaluations of closeness and interaction quality. Findings also suggest that autistic and NA individuals may evaluate interactions differently, with NA individuals placing greater value on their partner's intelligence and social presentation. Therefore, while these findings offer some promise that the benefits of AAT may extend beyond the laboratory (Jones et al., 2021) into real-world settings and increase social interest between autistic and NA adults, more systematic changes are likely needed to bridge the communicative and interactive divide between autistic and NA adults.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by The University of Texas at Dallas Institutional Review Board. The patients/participants provided their written informed consent to participate in this study.

KM, DJ, and NS led the design of the study, with contributions from AP and RA. DJ gathered and organized the data, with contributions from KM and KD. DJ conducted data analysis in collaboration with RA. DJ interpreted the data, with help from RA, AP, KD, and NS. DJ and NS wrote the manuscript, with contributions from AP, RA, and KD. All authors contributed to the article and approved the submitted version.

This research was funded by the Texas Higher Education Coordinating Board's Autism Grant Program.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.