Authors: (1) Yi-Ling Chung, The Alan Turing Institute (ychung@turing.ac.uk); (2) Gavin Abercrombie, The Interaction Lab, Heriot-Watt University (g.abercrombie@hw.ac.uk); (3) Florence Enock, The Alan Turing Institute (fenock@turing.ac.uk); (4) Jonathan Bright, The Alan Turing Institute (jbright@turing.ac.uk); (5) Verena Rieser, The Interaction Lab, Heriot-Watt University and now at Google DeepMind (v.t.rieser@hw.ac.uk). Authors: Authors: (1) Yi-Ling Chung, The Alan Turing Institute (ychung@turing.ac.uk); (2) Gavin Abercrombie, The Interaction Lab, Heriot-Watt University (g.abercrombie@hw.ac.uk); (3) Florence Enock, The Alan Turing Institute (fenock@turing.ac.uk); (4) Jonathan Bright, The Alan Turing Institute (jbright@turing.ac.uk); (5) Verena Rieser, The Interaction Lab, Heriot-Watt University and now at Google DeepMind (v.t.rieser@hw.ac.uk). Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 Background 2 Background 3 Review Methodology 3 Review Methodology 4 Defining counterspeech 4 Defining counterspeech 4.1 Classifying counterspeech 4.1 Classifying counterspeech 5 The Impact of Counterspeech 5 The Impact of Counterspeech 6 Computational Approaches to Counterspeech and 6.1 Counterspeech Datasets 6 Computational Approaches to Counterspeech and 6.1 Counterspeech Datasets 6.2 Approaches to Counterspeech Detection and 6.3 Approaches to Counterspeech Generation 6.2 Approaches to Counterspeech Detection and 6.3 Approaches to Counterspeech Generation 7 Future Perspectives 7 Future Perspectives 8 Conclusion, Acknowledgements, and References 8 Conclusion, Acknowledgements, and References 2 Background Interest in investigating the social and computational aspects of counterspeech has grown considerably in the past five years. However, while extant work reviews the impact of counterspeech on hate mitigation (Saltman and Russell, 2014; Carthy et al., 2020; Buerger, 2021a), none have systematically addressed this issue in combination with computational studies in order to synthesise social scientific insights and discuss the potential role of automated methods in reducing harms. Carthy et al. (2020) present a focused (2016-2018) systematic review of research into the impact of counter-narratives on prevention of violent radicalisation. They categorise the techniques employed in counter-narratives into four groups: (1) counter-stereotypical exemplars (challenging stereotypes, social schema or moral exemplars), (2) persuasion (e.g., through role-playing and emotion inducement), (3) inoculation (proactively reinforcing resistance to attitude change or persuasion), and (4) alternative accounts (disrupting false beliefs by offering different perspectives of events). The measurements of counternarrative interventions are based on (1) intent of violent behaviour, (2) perceived symbolic/realistic group threat (e.g., perception of an out-group as dangerous), and (3) in-group favouritism/out-group hostility (e.g., level of trust, confidence, discomfort and forgiveness towards out-groups). They argue that counter-narratives show promise in reducing violent radicalisation, while its effects vary across techniques, with counter-stereotypical exemplars, inoculation and alternative accounts demonstrating the most noticeable outcomes. Buerger (2021a) reviews the research into the effectiveness of counterspeech, attempting to categorise different forms of counterspeech, summarise the source of influences in abusive/positive behaviour change, and elucidate the reasons which drive strangers to intervene in cyberbullying. Here, the impact of counterspeech is mostly evaluated by the people involved in hateful discussions, including hateful speakers, audiences, and counterspeakers. In comparison, we focus on what makes counterspeech effective by comprehensively examining its use based on aspects such as strategies, audience and evaluation On the computational side, some work reviews the use of counterspeech in social media using natural language processing, including work outlining counterspeech datasets (Adak et al., 2022; Alsagheer et al., 2022), discussing automated approaches to counterspeech classification (Alsagheer et al., 2022) and generation (Chaudhary et al., 2021; Alsagheer et al., 2022), and work focusing on system evaluation (Alsagheer et al., 2022). However, work from computer sciences is not typically informed by important insights from the social sciences, including the key roles of intergroup dynamics, the social context in which counterspeech is employed, and the mode of persuasion by which counterspeech operates. Taking an interdisciplinary approach, we join work from the computer and social sciences. This paper is available on arxiv under CC BY-SA 4.0 DEED license. This paper is available on arxiv under CC BY-SA 4.0 DEED license. available on arxiv
