Today I am again at the Association of Internet Researchers AoIR 2016 Conference in Berlin. Yesterday we had workshops, today the conference kicks off properly. Follow the tweets at: #aoir2016.
As usual this is a liveblog so all comments and corrections are very much welcomed.Â
Platform Studies: The Rules of Engagement (Chair: Jean Burgess, QUT)
How affordances arise through relations between platforms, their different types of users, and what they do to the technology – Taina Bucher (University of Copenhagen) and Anne Helmond (University of Amsterdam)
Taina: Hearts on Twitter:Â In 2015 Twitter moved from stars to hearts, changing the affordances of the platform. They stated that they wanted to make the platform more accessible to new users, but that impacted on existing users.
Today we are going to talk about conceptualising affordances. In it’s original meaning an affordance is conceived of as a relational property (Gibson). For Norman perceived affordances were more the concern – thinking about how objects can exhibit or constrain particular actions. Affordances are not just the visual clues or possibilities, but can be felt. Gaver talks about these technology affordances. There are also social affordances – talked about my many – mainly about how poor technological affordances have impact on societies. It is mainly about impact of technology and how it can contain and constrain sociality. And finally we have communicative affordances (Hutchby), how technological affordances impact on communities and communications of practices.
So, what about platform changes? If we think about design affordances, we can see that there are different ways to understand this.Â The official reason for the design was given as about the audience, affording sociality of community and practices.
Affordances continues to play an important role in media and social media research. They tend to be conceptualised as either high-level or low-level affordances, with ontological and epistemological differences:
- High: affordance in the relation – actions enabled or constrained
- Low: affordance in the technical features of the user interface – reference to Gibson but they vary in where and when affordances are seen, and what features are supposed to enable or constrain.
Anne: We want to now turn to platform-sensitive approach, expanding the notion of the user –> different types of platform users, end-users, developers, researchers and advertisers – there is a real diversity of users and user needs and experiences here (see Gillespie on platforms. So, in the case of Twitter there are many users and many agendas – and multiple interfaces. Platforms are dynamic environments – and that differentiates social media platforms from Gibson’s environmental platforms. Computational systems driving media platforms are different, social media platforms adjust interfaces to their users through personalisation, A/B testing, algorithmically organised (e.g. Twitter recommending people to follow based on interests and actions).
In order to take a relational view of affordances, and do that justice, we also need to understand what users afford to the platforms – as they contribute, create content, provide data that enables to use and development and income (through advertisers) for the platform. Returning to Twitter… The platform affords different things for different people
Taking medium-specificity of platforms into account we can revisit earlier conceptions of affordance and critically analyse how they may be employed or translated to platform environments. Platform users are diverse and multiple, and relationships are multidirectional, with users contributing back to the platform. And those different users have different agendas around affordances – and in our Twitter case study, for instance, that includes developers and advertisers, users who are interested in affordances to measure user engagement.
How the social media APIs that scholars so often use for research areâ€”for commercial reasonsâ€”skewed positively toward â€˜connectionâ€™ and thus make it difficult to understand practices of â€˜disconnectionâ€™ – Nicolas John (Hebrew University of Israel) andÂ Asaf NissenbaumÂ (Hebrew University of Israel)
Consider this… On Facebook…If you add someone as a friend they are notified. If you unfriend them, they do not. If you post something you see it in your feed, if you delete it it is not broadcast. They have a page called World of Friends – they don’t have one called World of Enemies. And Facebook does not take kindly to app creators who seek to surface unfriending and removal of content. And Facebook is, like other social media platforms, therefore significantly biased towards positive friending and sharing actions. And that has implications for norms and for our research in these spaces.
One of our key questions here is what can’t we know about
Agnotology is defined as the study of ignorance. Robert Proctor talks about this in three terms: native state – childhood for instance; strategic ploy – e.g. the tobacco industry on health for years; lost realm – the knowledge that we cease to hold, that we loose.
I won’t go into detail on critiques of APIs for social science research, but as an overview the main critiques are:
- APIs are restrictive – they can cost money, we are limited to a percentage of the whole – Burgess and Bruns 2015; Bucher 2013; Bruns 2013; Driscoll and Walker
- APIs are opaque
- APIs can change with little notice (and do)
- Omitted data – Baym 2013 – now our point is that these platforms collect this data but do not share it.
- Bias to present – boyd and Crawford 2012
Asaf: Our methodology was to look at some of the most popular social media spaces and their APIs. We were were looking at connectivity in these spaces – liking, sharing, etc. And we also looked for the opposite traits – unliking, deletion, etc. We found that social media had very little data, if any, on “negative” traits – and we’ll look at this across three areas:Â other people and their content; me and my content; commercial users and their crowds.
Other people and their content – APIs tend to supply basic connectivity – friends/following, grouping, likes. Almost no historical content – except Facebook which shares when a user has liked a page. Current state only – disconnections are not accounted for. There is a reason to not know this data – privacy concerns perhaps – but that doesn’t explain my not being able to find this sort of information about my own profile.
Me and my content – negative traits and actions are hidden even from ourselves. Success is measured – likes and sharin, of you or by you. Decline is not – disconnections are lost connections… except on Twitter where you can see analytics of followers – but no names there, and not in the API. So we are losing who we once were but are not anymore. Social network sites do not see fit to share information over time… Lacking disconnection data is an idealogical and commercial issue.
Commercial users and their crowds – these users can see much more of their histories, and the negative actions online. They have a different regime of access in many cases, with the ups and downs revealed – though you may need to pay for access. Negative feedback receives special attention. Facebook offers the most detailed information on usage – including blocking and unliking information. Customers know more than users, or Pages vs. Groups.
Nicholas: So, implications. From what Asaf has shared shows the risk for API-based research… Where researchers’ work may be shaped by the affordances of the API being used. Any attempt to capture negative actions – unlikes, choices to leave or unfriend. If we can’t use APIs to measure social media phenomena, we have to use other means. So, unfriending is understood through surveys – time consuming and problematic. And that can put you off exploring these spaces – it limits research. The advertiser-friends user experience distorts the space – it’s like the stock market only reporting the rises except for a few super wealthy users who get the full picture.
A biography of Twitter (a story told through the intertwined stories of its key features and the social norms that give them meaning, drawing on archival material and oral history interviews with users) – Jean Burgess (Queensland University of Technology) and Nancy Baym (Microsoft Research)