Although we are talking about the automated production of memory in this book, these systems are still anchored by classification systems that open them up to a much longer held and well-established, as Foucault (2002) put it, order of things. It is also important to note that ‘The Taxonomy of Memory Themes’ discussed in Chapter Two served as the ‘ground truth’ (Amoore, 2020), so to speak, for the development of Facebook Memories. Established prior to its development, the memory classifications generated by Facebook's research studies were fed into the design of Facebook's current throwback feature. This was effectively a moment in which the formalization of a computational problem occurred and where there was an attempt to render the indeterminable and contingent into something calculable (see Fazi, 2018). Once this taxonomy of memories was in place, it provided the ranking algorithms with a clear-cut computational problem to ‘solve’ and optimize: what to surface, to whom and when. In other words, once there was a system in place for classifying memories within the taxonomy, the system had to then decide which memory, from all these many classified memories, should be targeted at the intended recipient and when they should receive it. Once the classificatory system is active within this social media archive, the focus then has to shift to retrieval and to the way in which this retrieval is instantiated in processes of ranking. Bringing memories to the surface requires, in this logic, a system by which they can be ranked – memories ranked at a certain level are the ones that then become visible. It is this ranking of memory that this chapter deals with.

Feedback loops and the surfacing of memories

In a Facebook Research report titled ‘Engineering for nostalgia: building a personalized “On This Day” experience’, Manohar Paluri and Omid Aziz (2016) outline the software engineering side to building the earlier iteration of Facebook Memories called On This Day. The claim behind this, they explain, is that they ‘wanted to make sure On This Day shows people the memories they most likely want to see and share, especially when it comes to the memories they see in News Feed’ (Paluri & Aziz, 2016).

Even something as intimate and personal as memory cannot escape the reach of social media and their datafied and circulatory logic. In this book we have explored the underlying processes that enable the selection and targeting of past content in the form of repackaged ‘memories’. Here we have highlighted the way that classification and ranking operate together to enable memories to resurface on social media throwback features. Through the combination of classification and ranking, the automated production and delivery of socalled ‘memories’ means that social media users do not need to dig; they are not excavating, as Walter Benjamin suggested, but instead that excavation is being done on their behalf. Benjamin noted that memories were always a way of mediating the masses of past experiences; this has not changed. These automated systems of social media remediate those memories through the classificatory systems that group them and then prioritize them, making them visible or invisible to us, and shaping how individuals and groups participate in those memories. Because, as Benjamin pointed out, memories have always been a mediation of the past, they can readily be reworked by these automated systems. As we have seen though, one problem with the automatic production of memory is authenticity. It is the act of producing memories that lends them authenticity; if that work becomes automated then potential tensions emerge around the legitimacy of that memory.

‘The promise of automation’, writes Mark Andrejevic (2020: 13), ‘is to encode the social so that it can be offloaded onto machines.’ In order to see the consequences this will have for memory and remembering, we suggest that there is a need to better understand the underlying classification and prioritization processes, what they are intended to do, as well as what implications and outcomes they have for people in everyday life. As a result, this book has sought to make a specific intervention into the automatic production of memory. Our contribution here has been to examine the role played by classification and ranking within these processes of automation. Once memories are opened up to classification and ranking, then the memories themselves will change, but so too will our understanding of what memories are. The concept of memory is unlikely to go untouched by these developments – indeed, we have sought to foreground the tensions that these processes of redefinition are already creating through features such as Facebook Memory.

The kind of automatic production and targeting of memories that we have described in the previous chapters is still relatively new. Yet it is already widespread and deeply embedded in how people relate to their past through social media content. As we have shown, processes of classification and ranking are central to how people encounter past social media content as memories. What this will mean for collective and individual memory will take some time to fully understand. However, in this chapter we would like to turn to a project that was recently completed by the first-named author in order to begin to think through and explore what these changes might mean, examining how people might come to respond and react to these packaged and targeted memories. The previous chapters showcase how the memorable is partitioned and promoted. In this chapter, we will reflect more directly on the reception of the classified and ranked memories with which users are presented. Given the scope of the issues, this is not a complete endeavour, but it begins to give glimpses into the variegated reception of automatically sorted memories that might then be pursued further. It will indicate the types of direction that memory making may be taking in the context of social media and mobile devices. In short, this chapter begins to explore something that is well- established but little understood as of yet. As discussed in Chapter One, we may know some of what happens when digital memories or mediated memories become integrated, but this particular chapter is about how people react to targeted memories. Partitioning and promoting the memorable through processes of classification and ranking assumes that the memory categories produced are fixed and distinct (Mackenzie, 2015). Yet, as we shall show, the processes of classification and ranking do not necessarily mean that memories fit neatly into those fixed grids of Facebook's taxonomy; nor are the reactions entirely in keeping with those imagined in the rhetorical ideals of the social media providers and coders. As this chapter shows, the reception of targeted memories in everyday life emphasizes the various nuances and tensions generated by the dual process of classification and ranking.