Abstract Social media are often accused of worsening the quality of online debate. In this paper, we focus on group polarization in the context of social media-enabled interaction, a dysfunctional group dynamic by which participants become more extreme in their initial position on an issue. Through a systematic literature review, we identified a corpus of 121 research papers investigating polarization in social media and other online conversational platforms and reviewed the main empirical findings, as well as theoretical and methodological approaches. We use this knowledge base to assess some recurrent accusations against social media in terms of their supposed tendency to worsen online debate. Our analysis shows that, while some concerns have been exaggerated, social media do contribute to increase polarization either by amplifying and escalating social processes that also occur offline or in specific ways enabled by their design affordances, which also make these platforms prone to manipulation. We argue against suggestions aimed at reducing freedom of speech in cyberspace and identify in inadequate regulation and lack of ethical design as the leading causes of social media-enabled group dysfunctions, highlighting research areas that can support the creation of higher quality online discursive spaces.