An Algorithmic Conception

An exploration of the suspicion and self-judgement that can accompany personalised online content

By Ella Relf, PhD Candidate, School of Culture and Communications
CAIDE Summer Research Academy 

Online targeted advertisements can be more than just ‘relevant’ or ‘creepy’; they can also place users in uncomfortable positions. To explore the unnerving suspicion and flickers of self-judgement that these encounters provoke, I want to describe a personal example. I am a young woman, and I am frequently shown advertising for pregnancy tests.

Pregnancy test 2

Anecdotally, a wide spread of female friends and peers have noticed the same, and hence I can somewhat simply understand this content as targeted towards me because I am female and within childbearing age. At the same time however, I cannot help but feel a layer of wariness towards these ads. As well as implicating my rough age range and gender, previous experience with relevant online content has led me to understand browsing as a highly targeted activity; I feel the system must have surely gathered more specific information like education, occupation, ethnicity, socio-economic class, residential area, purchase interests. At least some of this data, I imagine, would point to the fact that that I am not currently in want of children. I am conscious that other advertisers have capitalised on such details of my identity through their own marketing, and have no reason to doubt that the pregnancy test marketers could do the same.

Pregnancy test marketing must reach me then, at least sometimes, via a path which ‘knows’ that a test would not result in an exclamation of joy. Yet despite the fact that a proportion of women buy pregnancy tests in the hope of negative, through all my browsing I have never seen a pregnancy test advertisement targeted towards me that presented “not being pregnant” as the desired outcome. (Admittedly, that might be a more challenging marketing pill for society to swallow.) While over time the advertising I see has become more and more personalised and specific, the pregnancy test video pre-rolls have remained a relatively consistent piece of marketing content.  Knowing this, therefore, I am forced to confront three interpretive possibilities.

  1. That marketers have mistakenly concluded from my data that I would currently like to get pregnant. This positions the algorithmic judgement as off-the-mark.

    This is at once the most comforting option and also the most unnerving; I must then ask, as others have before, what is it about my tracked online behaviour that has led me to seem like I should consume this? (in this case, a pregnancy test?) (Cohn 2019). It might also lead me to scrutinise my browsing habits trying to identify these ‘markers’ that have placed me in the same category as women who “want” to become pregnant.

  2. The algorithmic system has an accurate grasp of my identity, and proceeds with the ad. In other words, it ‘knows’ I do not wish to be pregnant, but continues to send me this stream of happy, sentimental video pre-rolls.

    In this case, the apparently ‘mismatched’ or erroneous marketing must be merely aimed to seem that way. This seems aimed to remind me that ‘other’ women (whom the ad is for) desire a baby, leaving me to distance myself from their starry-eyed wish, and simultaneously cast a doubtful eye upon my own contraception methods. This now-deliberate act of “mis”-targeting rather nudges me to buy a pregnancy test to make damn sure it’s negative. Note here that the advertisement changes in meaning purely through my judgements about personalised targeting, rather than media content itself. This leads me to a third possibility.

  3. The algorithm knows better. In this case, either

    1. I “secretly do” (unbeknownst to me) wish to become pregnant, which the algorithm may be proposing through judgements gathered from my online behaviour, or,
    2. my online behaviour has left ‘traces’ (also unbeknownst to me) up that I am already pregnant.

    The latter is certainly not out of the question; multi-billion-dollar homewares and clothing store Target was enfolded into this exact scandal in 2012, when The New York Times reported on their targeting efforts revealed the use of highly detailed behavioural data about women at different stages of pregnancy (Duhigg, 2012). So accurate were their methods that they marketed early-pregnancy products to women before the women themselves knew they were pregnant. Although Target have since reconsidered the ethics of accurate marketing in this department, there is no doubt that marketers across the board are pressing closer and closer towards showing consumers content before it becomes an articulated, or even conscious, desire (John, Kim and Barasz, 2018: 64). Since then, research has shown that customisation and targeting practices which go too far can provoke consumer backlash over violations of privacy and excessive use of personal information (ibid, 63). For me, the idea that an algorithm is making these sorts of drastic judgements about my intentions or my body of which I am not even aware seems both intimately invasive and absurd.


    This dilemma positions me, the user, as never able to judge exactly how targeted content is when it reaches my screen, and from what personal information these content suggestions are made. In algorithmic culture, this personalisation paradox is ubiquitous. The opacity of targeting categories and the wide variety of content deemed ‘relevant’ ensures it is never clear how ‘personalised’ the content is intended to be, opening successive hermeneutics of suspicion. I am also left to confront advertisements with differing degrees of reflexivity - so if I spent an afternoon conducting online searches of pregnancy symptoms, for example, a barrage of ‘targeted’ ads would seem to have used simple tracking strategies, and feel minimally threatening. If instead, the same pregnancy ads emerge from my ‘normal’ browsing activities, they seem a leering presumption of my internet use and personhood.

    Of course, some targeting is more easily disregarded than others. The possibility of a deliberately-mismatched pregnancy ad carries enough weight for me, however, that I am sometimes inclined to regard both the company and myself with disgust; first for being advertised to in this fake-happy manner; and also for being tempted to doubt that I’ve done enough to prevent a real pregnancy. Even the wish to not be pregnant (in its absence) casts the user as failing to want what those women desire – and, dare I say, works to conflate abortion, safe sex and the unsexed female body together with whatever shadows reside at the opposing end of these precious, hopeful women on screen.

    Ella is a PhD Candidate in the Media & Communications department at the School of Culture and Communication and a member of the CAIDE Summer Research Academy. Her research is focused on the way digital environments influence identity choices; her interests include algorithmic and data capturing practices, cultural theory, ethics, and modes of online communication.

    REFERENCES

    Cohn, Jonathan. 2019. “My TiVo Thinks I’m Gay: Algorithmic Culture and Its Discontents.” TELEVISION & NEW MEDIA 17 (8): 675–90. https://doi.org/10.1177/1527476416644978.

    Duhigg, Charles. 2012. “How Companies Learn Your Secrets.” New York Times, February 16, 2012. https://www.nytimes.com/2012/02/19/magazine/shopping-habits.html.

    John, Leslie, Tami Kim, and Kate Barasz. 2018. “Ads That Don’t Overstep.” Harvard Business Review 96 (1): 62–69