Notes on: Matamoros-Fernandez, A & Farkas, J. ( 2021).  Racism, Hate speech and Social Media: I Systematic Review and Critique. Television and New Media. 22 (2): 205 – 224 DOI: 10.1177/1527476420982230 Dave Harris [a meta-study,focused on methodological problems really rather than themes, inevitably so. There might be more to come from these people?] They examined hundred and four articles with three research questions: which geographical contexts platforms and methods should be used? To what extent can we draw on 'critical race perspectives? What are the primary methodological and ethical challenges?. They found problems. Humans and technology continually interact to change 'sociality' and social media companies play an important role, especially in 'mediating and amplifying old and new forms of abuse, hate and discrimination' (206). Social media research has become a subfield with its own journal [Social Media and Society] and several special editions. There are both new and old racist practices. These include 'the weaponisation of memes', the development of fake identities, toxic subcultures, the spread of 'reactionary right racist influencers' and coordinated harassment [applied to Reddit, YouTube and Twitter]. Others have also looked at emoji and GIFs. Micro-aggressions and overt discrimination appear 'in platform governance and designs'. Some filters have enabled White people to appear in '"digital Blackface"',to lighten the skin of Black people, or to exclude users with specific ethnic affinities [on behalf of marketers or sometimes political groups]. They also moderate content in various ways. However they also enable anonymous harassment and humour which can display racial content. This article presents a literature review and critique of recent work. They use 'critical race perspectives' [CRP] to include CRT… intersectionality… Whiteness studies… Postcolonial theory… And critical indigenous studies' (207). These help analyse power relations and 'avoid perpetuating power imbalances… Since they ground ethical research and best practices in the experiences of marginalised groups… Standpoint theory'. They have been used since the early days of Internet research, much of it relying on Omi and Winant to see how 'racial categories are created and contested online'. Others have argued for more critical understanding of Whiteness, and pointed out that O and W focus too much on the state and on the US. A new interest in indigenous people such as Torres Strait Islanders has discussed struggle with opportunities and challenges [Carlson and Dreher 2018, Reynauld et al. 2018] but this is still scarce. Critical perspectives need more development Their sample of literature to be studied was gathered through Google scholar and the Web of Science and they chose scholarly articles in English, which were primarily focused on racism, hate speech, social media platforms and of recent publication [2014 – 2018] they try to include some key books as well but realise that longer and in-depth perspectives might be under represented. They used 'deductive content analysis followed by an in-depth qualitative analysis' (208). They coded a random subsample first, compared results and revised the approach, using variables such as year of publication, social media platform, methodological approach used in the study and whether the article had a positionality statement or used a particular CRP. Then they coded the methods and identified nine in total. Then they pursued an 'in-depth, open-ended qualitative analysis noting methodological and ethical challenges described in the literature… Grouped under overall themes and topics' (209). They acknowledge their own position as 'White, cis gender, heterosexual, European, middle-class researchers' and this help them analyse 'how Whiteness inform social media studies on racism' and thus to avoid reproducing racism themselves. Findings include the limited context for scholarship on racism and hate speech, mostly in the USA and then Europe, with a wide discrepancy between the Global North and Global South. The research is westernised. Twitter has been the most studied platform followed by Facebook, Reddit, Whispr,YikYak, Tumblr, Instagram and Tinder. WhatsApp and We Chat have not been studied. Twitter is relatively open and accessible. Both qualitative and quantitative approaches have been used more or less equally, with only a minority using mixed approaches. Text analysis predominates rather than 'interactional forms' (211). Quantitative forms use content analysis, sometimes automated rather than network analysis. However, quantitative methods tend to focus on the concept of hate speech, while racism is studied more by qualitative methods, reflecting 'a terminological divide' (212) which might indicate different 'emphasis on structural, ideological and historical dimensions of racial oppression' with qualitative researchers. The second finding was that only 23% of articles develop CRP approaches, again with 'a clear divide between qualitative and quantitative' approaches [45% versus 5%]. Less than half looked at Whiteness or the colour line in Dubois. One study looked at Fanon, another looked at 'front stage racism'. O and W is more widely used. Material referring to critical indigenous studies is 'scarce but present', including some work on Australian racism. Some researchers identify and neglected potential for CRP. Positionality statements are relatively rare [7%]. The third finding displayed 'key commonalities in the methodological challenges' (213), including difficulties in defining terms like hate speech, the intersection of identities in single victims, which presented particular challenges for automated identifications — one solution was to develop probabilistic language models, or to adopt particular definitions of White privilege. There are limits to the data, for example just using one platform, inability to get at historical data or deleted content. Problems with identifying coded forms of hate speech turned on trying to understand how the communities themselves learned these codewords and used them. Generally there was a problem with the 'ambivalence of social media communication and context' (214) especially the role of 'humour, irony and play' which became a particular problem with things like 'meme culture'. Ethical problems arose with difficulties of avoiding amplification, for example not publicising particularly hateful sites, or over sharing material, intruding on privacy. These were more prominent in qualitative studies, although potentially just as important with digital data. The particular problems of studying far right groups were 'largely absent'. They draw conclusions about the intersectional nature of Whiteness, and the importance of the 'ethics of care and standpoint theories'(215). They point first to the need to study racism outside the USA and to include minorities 'such as the Roma'. Intersectionality is also not sufficiently acknowledged, particularly by quantitative researchers. Particular platforms especially Twitter are overrepresented, although other platforms might well be suspected as having a key role. Important features include enabling anonymity and 'algorithmically suggesting racist content'. Platforms also attract different demographics — Twitter apparently particularly attracts racial minorities in the US. They were surprised to find qualitative and quantitative methods 'close to equally represented' (216). However, there were 'striking differences in the conceptual vocabularies' which indicated 'a disciplinary split' and different emphases on history ideology and structures. They think the quantitative articles mostly focused on 'surface level detection of hate speech without deploying connections to wider systems of oppression and without engaging with critical scholarship… [Which]… Tend to reduce racism to just overt abusive expression'. There is also an excessive focus on written text, which tends to ignore visual content, although this is suspected to be increasingly important. Only a minority of studies used CRP, and where they did, it was mainly to examine texts as well as the experience of users, and less so to explain 'the implications of materiality' [? Relation to structures again I think]. There should be more emphasis on intersectionality as a framework to understand 'how capitalism (class), White supremacy (race), and hetero patriarchy (gender) reflect and structure social media designs and practices'. As it is, there is a risk of 'ideological investments in colourblindness' and a neglect of how power operates on social media. There are only two studies on the inadvertent racism of White people, although they may be picking up the lack of extended studies in books and monographs. Indigenous perspectives are largely missing and need to be foregrounded, for example in emphasising the importance of place, and its relationship to developments in AI, or the importance of kinship networks and how they might be used to domesticate machines[among indigenous groups] . There is a danger of romanticising indigenous knowledge, but it should be explored. There is a danger of malpractice with large amounts of public data available, especially with Twitter and research should 'avoid perpetuating historical processes of dispossession through nonconsensual data extraction from marginalised communities' (217). [How exactly?]. They should take care not to identify people by the use of the verbatim posts. Only one study asked Twitter users whether they could include tweets. Others argued that Twitter could not be considered to be a private space, or that obtaining informed consent was impractical. One team suggested that if data is scraped from websites, the result should be posted on the same website. So, overall, we need a broader range of research, it should be more scholarly and pay more attention to the structural nature of racism, and it should be more reflexive. The position relative the authors should be acknowledged. It should be clear about what terms we are researching, in particular not focus too much on hate speech. We need more CRP