Emotions on Twitter as crisis imprint in high-trust societies: Do ambient affiliations affect emotional expression during the pandemic?

During the Covid-19 crisis, citizens turned to Twitter for information seeking, emotional outlet and sense-making of the crisis, creating ad hoc social communities using crisis-specific hashtags. The theory of ambient affiliation posits that the use of hashtags upscales the call to affiliate with the values expressed in the tweet. Given the deep functional tie between values and emotions, hashtag use might further amplify certain emotions. While emotions in crises-hashtagged communities have been previously investigated, the hypothesis of amplification of emotions through hashtag use has not yet been tested. We investigate such effect during the Covid-19 crisis in a scenario of high-trust Nordic societies, focusing on non-hashtagged, crisis hashtagged (e.g., '#Covid-19') and threat hashtagged (e.g., '#misinformation') tweets. To do so we apply XLM-RoBERTa to estimate Anger, Fear, Sadness, Disgust, Joy and Optimism. Our results revealed that crisis-hashtagged (#Covid-19) tweets expressed more negative emotions (Anger, Fear, Disgust and Sadness) and less positive emotions (Optimism and Joy) than non-hashtagged Covid-19 tweets for all countries except Finland. Threat tweets (#misinformation) expressed even more negative emotions (Anger, Fear, Disgust) and less positive emotions (Optimism and Joy) than #Covid-19 tweets, with a particularly large effect for Anger. Our findings provide useful context for previous research on collective emotions during crises, as most Twitter content is not hashtagged, and given the faster spread of emotionally charged content, further support the special focus on specific ad hoc communities for crisis and threat management and monitoring.

Vocal markers of autism: Assessing the generalizability of machine learning models.

Machine learning (ML) approaches show increasing promise in their ability to identify vocal markers of autism. Nonetheless, it is unclear to what extent such markers generalize to new speech samples collected, for example, using a different speech task or in a different language. In this paper, we systematically assess the generalizability of ML findings across a variety of contexts. We train promising published ML models of vocal markers of autism on novel cross-linguistic datasets following a rigorous pipeline to minimize overfitting, including cross-validated training and ensemble models. We test the generalizability of the models by testing them on (i) different participants from the same study, performing the same task; (ii) the same participants, performing a different (but similar) task; (iii) a different study with participants speaking a different language, performing the same type of task. While model performance is similar to previously published findings when trained and tested on data from the same study (out-of-sample performance), there is considerable variance between studies. Crucially, the models do not generalize well to different, though similar, tasks and not at all to new languages. The ML pipeline is openly shared. Generalizability of ML models of vocal markers of autism is an issue. We outline three recommendations for strategies researchers could take to be more explicit about generalizability and improve it in future studies. LAY SUMMARY: Machine learning approaches promise to be able to identify autism from voice only. These models underestimate how diverse the contexts in which we speak are, how diverse the languages used are and how diverse autistic voices are. Machine learning approaches need to be more careful in defining their limits and generalizability.