Emotions in Digital Media: Emotional responses to online messages presented with and without emotion labels – Dataset by Mirja Ilves, Aleksi Syrjämäki, Thomas Olsson, Joel Kiskola, Poika Isokoski, Anna Rantasila, Gary Bente, and Veikko Surakka The data consist of subjective emotional responses to social media messages. The effects of emotion labels presented with social media messages on participants’ emotional responses to the messages were studied in two online experiments. The stimuli (20 negative, 20 neutral, 20 positive) were selected from the FinnSentiment corpus (Lindén et al., 2020). It is a dataset containing 27 000 messages from the Finnish social media site Suomi24. The messages have been annotated by three independent annotators in terms of their sentiment (negative, neutral, positive). In the first experiment, participants read messages with negative, neutral, and positive sentiment, which were shown with and without emotion labels. In the emotion labeling condition, the sentiment of the message (negative, neutral, positive) was written below the message in capital letters, in dark gray color. After viewing each stimulus, participants filled in two questionnaire items. First, they assessed the valence of their emotional experience when reading the message. Secondly, they assessed the (in)appropriateness of the content of the message. Both measurements were done with -4 … +4 scales, where negative scores indicated negative valence / high inappropriateness, and positive scores indicated positive valence / high appropriateness. In the second experiment, participants read messages with negative and positive sentiment, which were shown with emotion label, non-emotional label or without any label. The no labeling and emotion labeling conditions were identical to Experiment 1. In the non-emotion labeling condition, the word “comment” was written below the message in capital letters, in dark gray color. The valence measurement was identical to Experiment 1. In addition, the participants rated how likely they would be to respond to the message on a 9-point scale (very unlikely - very likely) (Fishman et al., 2020). In the instructions, participants were told to rate the likelihood of replying if they had seen the message on a real online forum in response to their own message. The included files are: - Emotion label strengthens emotions_Exp1.xlsx – Table with the participants' ages and genders, the ratings of valence and appropriateness of the content of the message. The second sheet contains the average ratings. - Emotion label strengthens emotions_Exp2.xlsx – Table with the participants' ages and genders, the ratings of valence and likelihood of responding. The second sheet contains the average ratings. References: Lindén, K., Jauhiainen, T., & Hardwick, S. (2020). FinnSentiment – A Finnish social media corpus for sentiment polarity annotation. ArXiv. doi:10.48550/arXiv.2012.02613 Fishman, J., Lushin, V., & Mandell, D. S. (2020). Predicting implementation: comparing validated measures of intention and assessing the role of motivation when designing behavioral interventions. Implementation science communications, 1(1), 1-10. doi:10.1186/s43058-020-00050-4

2023