photo filters and collective normalizations

Here Comes the New Algorithm!

Photo Filters and Collective Normalizations

by Ulrike Bergermann, translated into English by Alexandra Cox (Extracts)

From: Amateur Photography: From Bauhaus to Instagram.

photo filters and collective normalizations
© 2019 Kehrer Verlag Heidelberg Berlin, authors, artists and photographers [and translators]. ISBN 978-3-86828-964-0
photo filters and collective normalizations
To place the means of photographic production in the hands of amateurs—what an emancipatory program that was. Its realization is complete with the advent of digital media: Own a smartphone, and you have a camera; get connected and you can curate your own gallery on Instagram. When the philosopher Friedrich Nietzsche opined, “our writing equipment takes part in forming our thoughts”, (i) the influence that photography was having upon his work still appeared manageable. When a photographer asks himself or herself how the camera codetermines his or her pictures, at least viewfinder and shutter are held in his/her hands and the lens functions similarly to the eye. However, today it is up for debate whether digital components of current photography practices do not in fact ‘co-write our pictures’, over which we have no power, which are not in our hands and are not even in our cameras any longer. Whereas the Bauhaus photographer Werner Gräff used his programmatic treatise Here Comes the New Photographer! (ii) to call on camera manufacturers to relax technical standards in order to enable new forms of expression for professionals and other users, barely a century later the boundaries between career photographers and everybody else have faded into the background and technologies for taking photographs to a high technical standard are widespread. Above all, though, it is no longer meaningful to address digital images solely as individual objects, as Winfried Gerling, Susanne Holschbach, and Petra Löffler establish. (iii) This is because information on the location of the shot, the camera, and the chip generation, along with programs which the file will have run through once it has been edited, tagged, and uploaded to a platform, are an integral component of the image; on the platform, in turn, the photo is commented on, people are marked on it (potentially automatically), images are liked, forwarded, etc. Therefore, photography theory’s occupation now lies with “photographic practices”, and with “photos” in the historical regard only. These practices include the activities of the people taking photographs, but also those of the technologies. Today, photography is a matter of “distributed power to take action”, (iv) the components of which are to be considered individually in each case. Everyone is able take photographs, everyone is an amateur, or an artist, or a professional—in everyone’s case, the work tools have a say as well.
photo filters and collective normalizations
The outcomes of this distribution of roles among professional producers, consumers, and others are prosumers. “Prosumers’ activities are linked with the activities of the programs”: (v) The early form of democratizing a participative culture, in which the first generation of databases and social networks provided storage space and ordering systems, was followed by an increasing preformatting of photographic practices by the big businesses providing the platforms, as Susanne Holschbach demonstrates. For example, Instagram offers various filters and editing tools that are able to alter photos and are mostly used to ‘beautify’ the photographed persons—self-presentation is Instagram’s most common mode of use. (vi) Now, photos are displayed according to “relevance”, and what is relevant is decided by Instagram’s algorithm. For example, since March 2016 it is no longer possible to define the sorting oneself, chronologically for instance. The algorithm does the sorting, and it also does the eliminating.[…]

[…] […]
photo filters and collective normalizations
As Web 2.0 goes commercial, Instagram and the like are gaining access to private data—for advertising purposes, and also in order to use them for developing their algorithms. These are based more and more commonly on self-learning principles: For example, programs perform evaluations once they have statistically defined, among enormous data quantities, which connections have been utilized by users the most frequently. If, therefore, a portrait is marked “hot”, the algorithm learns to associate the corresponding features with the positive rating and will steer programs accordingly—with the result that, for instance, white people will be drawn on for photo filters as “good-looking” image templates disproportionately often and corresponding stereotypes will take root in the repetition. The debate surrounding “racist algorithms” demonstrates that collective dysconscious collaborations influence the conditions of possibility of images of blacks and images of whites. Back in the age of analog photography there was the “Shirley Card”, a color panel bearing the photograph of a white female model named Shirley Page, (vii) which was used in the 1940s/50s and beyond to calibrate photographic cameras and printers. (viii) Kodak’s Multiracial Shirley Card had barely become commonplace when digitization made its breakthrough in documentary technology. (ix)

[…] […]

In summer 2015, the focus of the debate was no longer on film cameras or the photoshopping of print magazines, but on the filters that are available on social networks for uploading photos. […]

[…] […]

Most famous of all is FaceApp of 2016, the development of which was undertaken by neural networks, self-learning interconnected programs. (x) Developed by a small Russian team, the photo editor provides a host of filters that add a background to the selfie or portrait photo and/or modify the face, add a ponytail or beard, make the face look younger or older, put on a smile, or change the gender of those portrayed (Fig. x x) (xi) —editing the image in line with statistical knowledge concerning changes in faces, without causing the latter to lose their photo-realistic qualities. […]
photo filters and collective normalizations
[…] […]

[…] Image editing apps, working with neural networks, have already performed small-scale color changes before, (xii) but the more extensive use of artificial intelligence alters the images more comprehensively and realistically.

Are these still filters? They certainly do not imitate the working of analog filters. Here, nothing is moved in front of or on top of the portrait, either during the development process or in the finished product. Image data are filtered to suit certain vectors, in order to modify them in such a way as to make the product maximally comparable with the parameters of accordingly tagged data quantities. The first update included new photo filters (as they continue to be called) that stage a kind of race swap, “new photo filters that change the ethnic appearance of your face […] These new options, however, will likely cause some outrage: The filters are Asian, Black, Caucasian and Indian.” Following complaints, four months previously FaceApp had removed an app from the market which, among things, made skin color whiter beneath the symbol “hot”. Company chief Goncharov excused this racial bias with the explanation that the software had been trained based on an incorrect data pool, which included only pictures of white people, and in which the images had been marked “hot”.

[…] […]

As images circulate among various media environments, a photograph’s properties are not only to be grasped as mere possibilities in shifting contexts; as part of Big Data, data mining, and the corresponding new frameworks, they have transitioned into statistical probabilities which it is no longer possible to draw away from the real picture like a false veil. My interaction with photos on the internet will have codetermined the design and placement of images. Here Comes the New Algorithm and Here Comes the New Photographer, who is obliged to learn a new thinking in and against feedback-loops and echo chambers.
photo filters and collective normalizations
Ulrike Bergermann is a professor of media sciences at the Braunschweig University of Art. Her work concentrates on media theory, history of science, gender, and postcolonial studies.
photo filters and collective normalizations
(i) Friedrich Nietzsche (1882), in: Friedrich Nietzsche: Schreibmaschinentexte, eds. Stephan Günzel and Rüdiger Schmidt-Grépály (Weimar: Verlag der Bauhaus Universität, 2002), 18.
(ii) Werner Gräff, Es kommt der neue Fotograf! (Berlin: H. Rechendorf, 1929).
(iii) Winfried Gerling, Susanne Holschbach, and Petra Löffler, Bilder verteilen. Fotografische Praktiken in der digitalen Kultur, (Bielefeld: Digitale Gesellschaft, 2018). 10ff.
(iv)See Ilka Becker, Michael Cuntz, and Astrid Kusser (eds.), Unmenge. Wie verteilt sich Handlungsmacht?, (Munich: Fink Verlag, 2008).
(v) Susanne Holschbach, “Bilder teilen. Praktiken des Fotosharing”, in Bilder verteilen (see note 3), 17–80, here 38 and 45.
(vi) Ibid., 50. Means of editing aim at lightening, contrast enhancement, changing color temperature or depth of focus and also retro effects such as sepia toning, etc.; they have been available for iPhone since 2010, and since 2012 for Android as well.
(vii) Rosie Cima, “How Photography Was Optimized for White Skin Color”, Priceonomics, April 24, 2015; https://priceonomics.com/how-photography-was-optimized-for-white-skin/ [accessed January 3, 2019].
(viii) Lorna Roth, “Looking at Shirley, the Ultimate Norm: Colour Balance, Image Technologies, and Cognitive Equity”, Canadian Journal of Communication, vol. 43, no. 1, 2009, 111–136. photo filters and collective normalizations
(ix) Ibid., 122.
(x) faceapp.com – Andy, “Neural Networks: The Technology Behind FaceApp”, Masters of Media – blog of the New Media Digital Culture M.A., University of Amsterdam, September 24, 2017; https://mastersofmedia.hum.uva.nl/blog/2017/09/24/neural-networks-the-technology-behind-faceapp/ [accessed January 3, 2019].
(xi) Natasha Lomas, “FaceApp uses neural networks for photorealistic selfie tweaks”, TechCrunch, August 2, 2017; https://techcrunch.com/2017/02/08/faceapp-uses-neural-networks-for-photorealistic-selfie-tweaks/?guccounter=1 [accessed January 3, 2019].
(xii) Natasha Lomas, “Teleport’s Neural Networks Let You Try Before You Hair Dye”, TechCrunch, September 19, 2017; https://techcrunch.com/2017/08/09/teleports-neural-networks-let-you-try-before-you-hair-dye/ [accessed January 3, 2019].

Click on the poppy for Susanne Holsbach’s essay: