Speculative Diagrams: experiments in mapping Youtube - David Benqué
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Speculative Diagrams: experiments in mapping Youtube. Betti Marenko1, David Benqué2 Abstract: The notion that the future has a ‘shape’ is a 1Central Saint Martins Uni- deep-rooted construct, the cornerstone of how chance versity of the Arts Lon- is mediated, for example through the ‘distributions’ don, United Kingdom b.marenko@csm.arts.ac.uk of probability theory. Algorithmic prediction, via machine learning, builds on these shapes and ampli- 2 School of Communica- fies their complexity and authority. While the problem- tion, Royal College of Art, London, United Kingdom atic effects of this predictive regime and the preemp- tive politics it supports are objects of concern for scholars and practitioners across the humanities, social sciences, art, and philosophy, design is surpris- ingly disengaged from this conversation. Instead, it is either concerned with data visualization, often without questioning its positivist ontology, or with ‘seamless’ non-interfaces which effectively seek to remove choice. Against this backdrop, our proposal brings together design theory and design practice to inter rogate current modes of algorithmic prediction and the construction of subjectivity enabled by ‘choice design’. Our designed artifacts are diagrams to think through practice about the shape(s) of the possible. Rather than designing predictable futures, we aim to use diagram-making to expose and reframe choice design. These design artifacts - initial and ongoing experiments in mapping YouTube recommendations - are a series of computational diagrams that weave together the tools of computational prediction, critical design practice, and theory. Keywords: algorithmic prediction; diagrams; recommendation; criti- cal data visualization; 1
Frictions and Shifts in RTD #rtd2019 #researchthroughdesign #delft #rotterdam 2 3 Marenko, Benqué | Architectures of Choice Vol.1 Youtube.
Introduction 1. Algorithmic governmentality: Our collaboration between design theory and practice (Marenko and the calculative ontology of prediction Benque, 2018) investigates/ complicates/ criticises the practice and Every time a user is offered a series of recommendations generated by a politics of algorithmic prediction: future-building via machine learn- digital platform - Google, Facebook, Amazon or YouTube - they are being ing. It sets out to position diagramming as a computational count- worked on by a technology (ranking algorithms) that operates by ana- Frictions and Shifts in RTD er-practice, to reclaim the possible from algorithmic capture. We use lysing data on past behaviours to forecast future behaviours (Covington diagrams - speculative visualization mapping the unformed and the et al, 2016). However, predictive analytics is more than this. It orients unstable as they feed into changeable situations (O’Sullivan, 2016) - potential tendencies. What media theorist Mark Hansen describes as to think through, manipulate, and theorise prediction from a design “the calculative ontology of prediction” (Hansen, 2014, 38) – the im- standpoint. We use diagram-making as a practice that affords ways of pact of technomedia on the texture of human experience and on any probing the possible as it emerges through modes of prediction and subsequent behaviour – is a systemic way of designing tendencies and speculating on a future taken as fluid, rather a solid material. In this propensities at a precognitive level, in ways that are not fully accessi- paper we move from this starting point into our first practical case ble to conscious cognition and perceptual awareness. Our forthcoming study: a mapping of recommendations on the Youtube platform. behaviours are predicted not simply before they happen, but before we might even think of making them happen. Thus, the evaluation of Any interface to a catalogue of media is, explicitly or not, an order- how systems of choice are designed concerns also their role in build- ing of choices. While this is inescapable—there is no way to present ing identities - what philosopher Antoinette Rouvroy calls regimes of a list without some kind of ordering—it is also problematic. While algorithmic governmentality (Rouvroy, 2016). This is how recommen- user-friendly interfaces appear to be about offering choices, their dations work: when Youtube or Netflix suggest what to watch next, they underlying goal is to direct viewers “toward content more likely to not only build user profiles, but generate social identities drawn on a keep them engaged and subscribed” (Arnold, 2016, 99). Systems that multitude of fragmented points. In other words, prediction turns into aim to maximise user “engagement” (eyeballs capture for advertis- prescription. Classifiers do not extract categories from a neutral and ing) encourage to the continuous foregrounding of “fresh” (Coving- objective standpoint, but actively create them (Cheney-Lippold, 2011). ton et al, 2016) and “viral” content (Jiang et al, 2014)—this has been The granularity of captured user data is used to recursively construe shown to have hugely detrimental effects, from the traumatising of (and obscure) a live-feed of our every single subject position. By simul- children (Bridle, 2017) to the spreading of white nationalist ideol- taneously reflecting back and building up, this process is a continuous ogy (Lewis, 2018) . As “anticipatory design” (Shapiro, 2015) shows, modulation of the subject into recursive discrete data sets, endlessly the idea of choice design may result in the removal of options on the distributed on a multitude of access points - what Gilles Deleuze in his pretext of making decisions on users’ behalf for—allegedly—their Postscript on Control Societies calls the ‘dividual’ (Deleuze, 1995, 180). own benefit. Behind a rhetoric of convenience, the notion of reliev- ing users from “decision fatigue” masks a political project. Our con- 2. Mapping YouTube cern is not to challenge the need for an ordering system—with over The intricate architecture of YouTube recommendations, and the 400 hours uploaded every minute and 1.8 billion logged-in monthly challenges posed by its sheer volume of content, are manipulated by users on YouTube it is hard to argue that some interface is required engineering decisions concerning which type of content is promoted to parse this content. Instead we want to critically examine the pol- on the platform (Covington et al, 2016). This is a notoriously opaque itics embedded in the design of systems that predict “what to watch process (partly because of non-disclosure agreements) for the engi- next” and use this critique to underpin our design interventions. neers themselves who have only limited understanding of how the system predicts - let alone the public (Burrell, 2016). The secrecy sur- The designed artefacts proposed here – experiments in mapping You- rounding YouTube recommendation system, and its lack of account- Tube recommendations, titled Architectures of Choice Vol.1 Youtube – are ability, has prompted several initiatives to ‘map’ it from the outside: a programme of work toward a visual understanding of ‘choice design’. reverse engineering strategies (Albright, 2017), ‘manual’ techniques Their aim is twofold: to show how architectures of choice are being (Lewis, 2018), and most notably the AlgoTransparency project designed designed in our everyday, impacting on the construction of algorith- to expose the bias in recommendations after the 2016 US presiden- mic identities (Cheney-Lippold, 2011); and to draw attention to the tial election (Chaslot et al, 2016, Lewis and McCormick, 2018). complexities of (and negotiations needed for) studying them through a critical practice. These experiments are framed by three theoretical These examples seek to gather data/evidence in order to hypothesise on perspectives: the idea of algorithmic governmentality to foreground how the system might be producing detrimental effects. All three use net- the politics implicit in choice design (Rouvroy, 2016); ideas concerning work graphs, which model data as nodes (e.g. channels or individual vid- maps and diagrams as a way to extract methodological friction from Figure 1. Youtube “Up Next” eos) and relations (e.g. recommendations). While networks are well suited existent mapping systems (Deleuze and Guattari, 1988); a critique recommendations for the vid- to mapping the the complexity of such as vast recommendation system - eo “Saying Logan Paul 100,000 of design’s lack of engagement with these issues through an assess- Times” by MrBeast. Screen arguably the only available way to model such data - their claim to provid- ment of the positivist ontology of data visualization (Drucker, 2011). capture by the authors on ing any sense of ‘transparency’ does not scale well at all. The hand crafted 19.10.2018, url: https://www. graph of a small number of actors used by Lewis (2018) shows a complex youtube.com/watch?v=_FX6rml2Yjs web of co-appearances in videos and remains relatively readable. When attempting to map Youtube in its totality however, AlgoTransparency’s Youtube Map becomes nothing more than a “hairball” or “spaghetti bowl” - hardly revelatory about the processes behind them (Bounegru et al., 2017), in fact counter-producing a sort of visibility-rich opacity. Seeing, in this case, is not necessarily knowing (Ananny and Crawford, 2016). #rtd2019 #researchthroughdesign #delft #rotterdam 4 5
3. Designed artefacts: speculative diagrams for mapping YouTube recommendations a. Start on Youtube homepage Figure 2a, b, c. Three probe designs: a. ripple.py Visualisation We then query the graph database to retrieve the data from probe We approach the space of Youtube mapping by building on, and being mind- b. simple_digger.py sessions and visualise them. Python was used to experiment with ful of, these insights. The prototypes we show and discuss here use au- c. digger.py popular network visualisation packages networkx and matplotlib [fig. tomated techniques to collect, store and visualise recommendations as List recommended videos 5]. We also relied on the built in Neo4j browser during testing, and it Frictions and Shifts in RTD network data. They must be taken as explorative diagrams that attempt provided one of the more flexible ways to explore the data, a mixture to expose the logics of algorithmic capture. It must be clarified that the of automated layouts and manual manipulation [fig. 3]. We tested aim of our interventions is not to claim solutions to the shortcomings for each generating flowcharts for Mermaid.js [fig.4] but these quickly reached outlined above. Rather, the aim is to open up paths of enquiry that con- the limits of the library resulting in crashes and unreadable outputs. front the predictive apparatus and to keep on evidencing the challeng- Finally, we tested Gephi, an open source graph analysis and visualisa- es embedded in the tools through which this confrontation happens. tion software popular amongst Digital Humanities practitioners. [fig. 6]. Load video page Collection None of these tools were satisfactory. Beyond the structural as- Data are collected with a headless browser (Selenium) to simulate pect of simply having too many data points to show, simple things user interaction with web-pages and ‘click’ through recommenda- add video data like the lack of text-wrapping functionality to display video titles tions. This allows for quick prototyping using the browser’s develop- to graph on multiple lines of text rendered our outputs unreadable very er tools to identify parts of the page we are interested in collecting quickly. We addressed this by zooming in on the granularity of in- or interacting with. Starting from the YouTube homepage, with its dividual paths/traces and to design our own visualisation using list of new videos and topics, we follow links to video pages and re- D3.js. [fig 7 & 8]. Importantly, practical concerns for readability cursively list recommendations. Three different ‘probe’ designs List recommended videos highlighted the performative nature of data as material and the balance the breadth and depth of the mapping in different ways: challenges of data visualization - discussed in the next section. ripple.py [fig. 2.a.] attempts to follow all recommendations for a giv- en number of recursions. This exhaustive approach was very time Repeat for consuming, and the resulting visualisations (see below) became n levels for each unreadable after the first recursion. A drastically simpler approach is taken in simple_digger.py [fig.2.b.] which selects only one video at random in the recommendations list and repeats the process for a given number of times. This approach was also unsatisfactory, as it Load video page 2018-08-01 13:09:50 presented single paths which fail to capture the interconnected na- ture of the recommendations made evident by the previous exper- Figure 3. Ltd isp: M247 Visualisation in the Neo4j Browser interface. loc: England iments. Digger.py [fig.2.c.] is a compromise which captures all of the add video data type: digger (n=10) recommendations at each step before choosing one at random. to graph The collected data are stored in a graph database (Neo4j) where videos are represented by nodes, and recommendations by edges connecting them. Each session is logged with a time stamp and in- dicates the type of probe used. This database can then be queried to return particular sessions or sub-graphs for visualisation. b. Start on Youtube homepage c. Start on Youtube homepage List recommended videos List recommended videos Pick one video at random Pick one video for each at random Load video page Load video page Load video page add video data to graph add video data repeat n times to graph repeat n times #rtd2019 #researchthroughdesign #delft #rotterdam 6 7
Dua Lipa ‐ IDGAF (Official Music Video) Maroon 5 ‐ Girls Like You ft. Cardi B Charlie Puth ‐ 'How Long' [Official Video] Figure 5. Visualisation of two 2018-07-12 22:20:20 2018-07-13 14:29:51 ripple.py sessions using Python Jonas Blue ‐ Rise ft. Jack & Jack packages networkx and matplotlib. Shawn Mendes ‐ Treat You Better isp: Virgin Media a. shows twoisp: Virgin levels of Media recommen- dations from the Youtube loc: England loc: England Charlie Puth ‐ The Way I Am [Official Video] Charlie Puth Special Olympics Performance in Seattle | LIVE 7‐1‐18 homepage. type: ripple (n=2) b. shows thetype: first ripple level (n=1 + 1) Frictions and Shifts in RTD Charlie Puth ‐ Marvin Gaye ft. Meghan Trainor [Official Video] of recommendations followed by one random pick. Talk ‐ Why Don't We [Official Music Video] Charlie Puth ‐ One Call Away [Official Video] Dhadak ‐ Title Track | Dhadak | Ishaan & Janhvi | Ajay Gogavale & Shreya Ghoshal | Ajay‐Atul Gali Mein Aaj Chand Nikla | HD Señor Grinch | Rudy Mancuso, Lele Pons & King Bach Make Your Own Lipstick Balm & Eyeshadow Makeup DIY Craft Do It Yourself a. FilterCopy | If Salary Were A Person | Ft. Wamiqa Gabbi and Veer Rajwant Singh Shawn Mendes Destroys James In a Cover Battle #LateLateShawn Pehli Baar | Dhadak | Ishaan & Janhvi | Ajay Gogavale | Ajay‐Atul | Amitabh Bhattacharya Clean Bandit ‐ Solo feat. Demi Lovato [Official Video] Funny Musical Barber | Rudy Mancuso The Love Mashup ‐ Atif Aslam & Arijit Singh 2018 | By DJ RHN ROHAN | Is this love or pain ? Yad Lagla | Official Full Video Song (2016) Nagraj Popatrao Manjule SANJU: Ranbir Kapoor to Sanjay Dutt ‐ The Transformation | Rajkumar Hirani | In Cinemas Now Ishq Di Baajiyaan ‐ Lyric Video |Soorma |Diljit | Taapsee |Shankar Ehsaan Loy | Gulzar Kavi Kumar Azaad aka Dr. Haathi FUNERAL Full Video Phir Bhi Tumko Chaahunga ‐ Full Video | Half Girlfriend| Arjun K,Shraddha K | Arijit Singh| Mithoon Dhadak | Official Trailer | Janhvi Kapoor | Ishaan Khatter | Shashank Khaitan | 20 July Koi Vi Nahi (Full Video) | Shirley Setia | Gurnazar | Rajat Nagpal Latest Songs 2018 | Speed Records LØVELY HEART TOUCHING 2018 ❤ JUKEBOX 2018 | BEST ROMANTIC JUKEBOX ❤ | BOLLYWOOD ROMANTIC Raghav Juyal and Shakti dance performance in Dance Plus 3 b. kar har maidaan fateh re bandeya full lyrical video song | Ranbir Kapoor | Shreya Ghoshal 2018-07-12 22:20:20 2018-07-13 14:29:51 Figure 4. Visualisa- | Halwai Prank | By Nadir Ali In | P4 Pakao | 2018 tion using Mermaid.js isp: Virgin Media Marshmello & Anne‐Marie ‐ FRIENDS (Music Video) *OFFICIAL FRIENDZONE ANTHEM* isp: Virgin Media loc: England loc: England 4. The positivist ontology of data visualization type: ripple (n=2) type: ripple (n=1 + 1) This project is caught in a paradox - which is also at the core of the Talking Tom Shorts – Ultra Marathon (All Episodes) Dhadak ‐ Title Track | Lyrical | Dhadak | Ishaan & Janhvi |Ajay Gogavale & Shreya Ghoshal |Ajay‐Atul digital humanities (Kitchin, 2014): how are we to critique the epis- Secret Santa | Lele Pons temologies of data and algorithms, while at the same time relying MARQUESE SCOTT | SHAKTI MOHAN DJ SNAKE MAGENTA RIDDIM (EXTENDED VERSION) HRVY ‐ Personal (Official Video) on them for our research? Here we bring the theoretical distinction Tera Zikr ‐ Darshan Raval | Official Video ‐ Latest New Hit Song between maps and tracing to bear on the specifics of data visualisation. #1 Best of Armaan Malik 2018 | Top 10 Songs | Armaan Malik Latest Songs | Romantic Hindi | Top Hits Domino | Rudy Mancuso Circle Of Love | Rudy Mancuso Mapping concerns imagining, and experimenting with, forms of Top 15 Most Shocking Reactions To Red Card in Football world-building only partially embodied in, and expressed by, their cool kids don't cry music video Rudimental ‐ Toast To Our Differences (feat. Shungudzo, Protoje & Hak Baker) (Official Video) given visual representation (Deleuze and Guattari, 1988). As a tool New Hair, New Me | Anwar Jibawi & Rudy Mancuso that unfolds potential (Corner, 1999), mapping emphasizes a pro- 30 THINGS YOU KEEP DOING WRONG Logan Paul ‐ Help Me Help You ft. Why Don't We [Official Video] ductive future-facing capacity to visualize simultaneously what is Wiz Khalifa Takes a Lie Detector Test | Vanity Fair and what is not there yet. Conversely, tracing belongs to a logic of Rudimental & Major Lazer ‐ Let Me Live (feat. Anne‐Marie & Mr Eazi) [Official Video] reproduction. It copies from the same model, it is always identical to Anne‐Marie ‐ 2002 [Official Video] itself and therefore lacks genuine openness to change. This distinc- OneRepublic ‐ Start Again ft. Logic tion is helpful in articulating the pitfalls of using data visualisation to expose/critique algorithmic systems, as most data visualisations Rudimental ‐ These Days feat. Jess Glynne, Macklemore & Dan Caplen [Official Video] claim to be tracings—neutral and objective depictions of reality. APES**T ‐ THE CARTERS Cardi B, Bad Bunny & J Balvin ‐ I Like It [Official Music Video] Data visualisation has been critiqued as a process where decisions are made reflecting “assumptions, unstated (often unacknowledged) Rudimental ‐ Not Giving In ft. John Newman & Alex Clare [Official Video] ideological perspectives and subjective judgments” (Boehnert, 2016). THE TOP SEXIEST BLIND AUDITION || THE VOICE GLOBAL 2017 [HD] The choices embedded in data visualization concern not only what to include, and what to leave out, but also claims of objectivity and Childish Gambino ‐ This Is America (Official Video) neutrality made by what is effectively the result of (ideological) se- Calvin Harris, Dua Lipa ‐ One Kiss (Official Video) #rtd2019 #researchthroughdesign #delft #rotterdam Major Lazer ‐ Know No Better (feat. Travis Scott, Camila Cabello & Quavo)(Official Music Video) 8 9 Kitchen Nightmares US S02E08 Sabatiello's
ISP: UK-2 Limited - Location: England lections. “Most information visualizations are acts of interpretation masquerading as presentation (…) they are images that act as if they Youtube homepage Mike Tyson - All ALL 6 OF MIKE TOP 10 IDIOTS WHO Knockouts TYSON LOSSES CHALLENGED are just showing us what is, but in actuality, they are arguments made Performed by the PROFESSIONAL in graphical form” (Drucker, 2014). Boehnert uses the terms “digital Legend HD FIGHTERS - Part 2 positivism” (over-simplistic take on how data mediates knowledges), Frictions and Shifts in RTD “datawash” (data visualisation that obfuscates knowledge on contro- versial issues) and “darkdata”( what is not measured or datified may 10 Dec 2018 5 Oct 2015 15 Mar 2018 8,575,207 13,102,650 14,327,927 be more relevant than what is) - to zoom in on the political nature of This is interesting! Jordan Peterson Clips LISTaddiction data visualization practices and tools. Data visualization carries val- ue judgments that are patently obfuscated by the blanket of digital positivism (Boehnert, 2018). Modes of “knowledge visualisation” can be used instead, which declare their ongoing situatedness and par- Holy Shit! Moments Top 10 People The Funniest Off- 13 Times Spoiled in MMA #15 - 6 getting OWNED! The-Cuff Comebacks Kids Got OWNED By tiality and build on these to create awareness and stimulate critical EVER [ROUND 2] Parents insights. Drucker’s distinction between data “which are assumed to be given” and capta which are “actively taken” is a key demarcation here (Drucker, 2011). While algorithmic transparency - “designed to give audiences an appreciation of programming decisions and con- 16 Nov 2018 23 Dec 2018 15 Dec 2017 6 Apr 2018 4,797,494 15,986,342 5,154,076 4,370,852 sequences” (Ananny and Crawford, 2016, 5) might be seen by some as a goal, it is also fraught with limitations. What matters, rather Mixed Martial Edits Top Things MotleyTV Viral For Miles are “meanings achieved through relations, not revelations” (ibid.). Our last experiment [new fig.] shows how foregrounding the re- Footage of 14 5 Times People THE 5 BEST BIGFOOT Is this 'finally' lations between a collection of paths or traces through the recom- Mythical Creatures Accidentally VIDEOS EVER Bigfoot captured -13 14:29:51mendation system might offer new understanding. By acknowl- That SHOULDN'T Spotted Mythical RECORDED! (Not in in stunning Exist Creatures on Tape! any particular quality? June 4 edging the impossibility of a full map (problems of scalability), order) (Disclose Screen) WORK TRANSIT HOLDINGS LLC this experiment does not just suggest that knowledge about the as system can only be grasped at a micro-scale. Importantly, it re- 29 Mar 2018 19 Dec 2017 8 Nov 2018 7 Jun 2018 ple (n=1) lies on traces to build new understanding and a mode of knowl- Figure 6 ripple.py session 10,911 4,970,826 206,490 1,035,909 Disclose Screen 'The edge-making that is situated, incomplete and speculative. The RareDoomed Video: Cleveland Balloonfest Japan Tsunami Geographic America's small town in Canada John Kerry: Trump clearly doesn't understand of '86 | National America This visualised with Gephi. Astronaut Was All Alone In Outer Space, But T... Forecaster says Florence could bring flash flooding Viral For Miles Top 5s Finest Sasquatch Central Grimreefar' Why graphene hasn’t taken over the world...yet Flying into the eye of Hurricane Irma Bernstein: Trump presidency teetering with an unst... Terrifying FAMILY InMoments as Both Engines Failed on Appro... If This Doesn't Make You a Believer, I Doubt Anyth... Pilot describes flying through Hurricane Florence California Is Sinking — and Now Could Flood | KQED... 5 Nostradamus Predictions For The ROYAL ... A Two Mile Crack Has Opened Up In Arizona – And Ex... N.C. family sues HGTV show for "disastrous" home m... Storm simulator recreates Category 5 hurricane con... Televangelists: Last Week Tonight with John Oliver... Hurricane Florence: Forecast Track Moves Further N... Kaplan graduate says degree means "absolutely noth... Millions brace for Hurricane Florence Is WALMART RV PARKING Over? Why are Walmarts turni... They Are Here: Huge Alien Craft In Our Solar System European hurricane model shows Hurricane Florence ... Meet the Judge Who Went Viral For His Creative Pun... Hurricane hunter flies over Category 4 storm Flore... The Difference between the United Kingdom, Great B... The Gaffney Incident: The Night Police Talked to a... Hurricane FLORENCE & OLIVIA 2018 | News Update | W... BIGFOOT In The Bigfoot encounter Law Enforcement The Interactions We Finally Know What Killed the Aztecs! Anderson Cooper calls out Sarah Sanders' op-ed def... Tsunami Climbing: Incredible video of ship heading... Why the poorest county in West Virginia has faith ... This plane could cross the Atlantic in 3.5 hours. ... South Carolina Coast Evacuating Due to Hurricane F... Don Lemon: Trump is lying, even about good news Jim Cantore Hurricane Florence Could Be an 'East Coast Harvey' vs. Category 5 Winds Forest - Part 3 at the Oregon Officer shares Are Increasing See What Happens When A Plane Violates Presidentia... Rams vs. Raiders Week 1 Highlights | NFL 2018 Ships In Storms Video Compilation [REAL FOOTAGE - ... Shannon Sharpe was 'impressed' by Aaron Rodgers du... Hurricane Florence sparks mandatory evacuations in... When hurricanes collide: The Fujiwhara Effect Hurricane Katrina Day by Day | National Geographic Caves, Part 1 Story (August 2, (June 5, 2018) Seahawks vs. Broncos Week 1 Highlights | NFL 2018 Colin on Green Bay 'escaping' with a win vs Chicag... Greenberg: Serena Williams’ US Open final controve... El huracán Florence se acerca y trae mucha preocup... Racist, sexist or out of order: Serena Williams de... Latest forecast for Hurricane Florence Chinese Drywall Nightmare Southeastern U.S. braces for Hurricane Florence as... 2017) US Open Winner Naomi Osaka Speaks Out On Controver... Mike Tomlin calls out Stephen A. during interview ... Tropics Update: Tracking Hurricane Florence, other... Jimmy Kimmel's FULL INTERVIEW with President Georg... Longest Passing Plays in NFL History (95+ yards) Did Sexism Cost Serena Williams U.S. Open Title? |... Hurricane Florence strengthens to Category 4, targ... TRANSFER PRANK: NFL QB Jared Goff pranks unsuspect... Aaron Rodgers - The Bad Man Max Kellerman: ‘I’m the only one who doesn’t’ owe ... TOP 10 BEST TORNADOES UB: Weather update as of 6:04 a.m. (Sept. 10, 2018) Massive 'one ring' of black holes, neutron stars d... Bears vs. Packers Week 1 Highlights | NFL 2018 Stephen A. says Serena Williams was wrong for 2018... Klavan Loses It On Serena Losing It Why Hurricane Categories Make a Difference Stephen A. says Aaron Rodgers is the greatest quar... Stephen A. comes down on the Steelers after their ... Max Kellerman: Mack trade could backfire right awa... One-Hit Wonders You Had No Idea Are Dead Michael Vick on Tyrod's Week 1 performance against... 12 Dec 2018 Obama bashes Trump, takes credit for good economy 19 May 2014 3 Aug 2017 6 Jun 2018 Serena Williams & Sexism in Sports - Between the S... Stephen A. on Dallas Cowboys: ‘What can go wrong, ... Shocking US Open final as Serena Williams loses, b... Greg Jennings was 'beyond impressed' with Rodgers'... ct to Aaron Rodgers’ ... Jets vs. Lions Week 1 Highlights | NFL 2018 Chiefs vs. Chargers Week 1 Highlights | NFL 2018 Will Cain: Serena Williams drama is a 'disgrace' t... Serena Williams was fined $17,000 for arguing with... Stephen A., Max debate how Colin Kaepernick's ad i... 81,508 Sexism in tennis: Debate flares up after Williams ... some vines i think about a lot How NYC plans to survive the next superstorm 468,992 175,250 6,413 Shady Things About The Cast Of American Pickers Weird Things Everyone Ignores About Serena William... 2018 US Open Highlights: Serena Williams' dispute ... curb your memes to watch instead of sleeping Vines for when you’re alone at 3am TRY NOT TO LAUGH OR SMILE CHALLENGE *IMPOSSIBLE* Amigos (Ep. 2) | Lele Pons, Rudy Mancuso, Juanpa Z... The Bachelorette | Lele Pons Vines I cannot live without (RIP Vine
Trace: 2019-01-21 12:11:06 Type: simple_digger ISP: UK-2 Limited - Location: England Figure 8 simple_digger.py session visualised with D3.js. 5. Reflections via mapping and diagramming The mapping experiments presented above are diagrams of how choice Who wants to be a Thief Tries to Youtube homepage WOW! Maurizio design in YouTube recommendations constructs algorithmic subjects. Millionaire - Attack Jeremy Sarri slams his As digital media theorist Luciana Parisi argues, this subject is “neither Major Fraud After Being Chelsea players | solely an enslaved component of machines not its deluded interactive Frictions and Shifts in RTD Exposed | The "This is a huge, Jeremy Kyle Show huge gamble!" user. Instead the subject is being reconfigured from the standpoint of a learning machine” (Parisi, 2019). Put differently, the subject emerges 10 Apr 2008 13 Sep 2016 19 Jan 2019 from forms of predictive patterning which retrofits choice (and think- 9,168,308 2,801,724 525,120 ing itself) as determined by its searching into the unknown. Thus, any RomeijnLand The Jeremy Kyle Show BT Sport attempt at mapping choice ought to acknowledge these mechanisms. For this reason we consciously repurpose computational tools as “cog- nitive prostheses” (Srnicek, 2012) to provide purchase on otherwise ungraspable systems. Such acts of building mediators between the Michael Gove rips Jacob Rees-Mogg Ant & Dec's 20 Police Officers into Jeremy Corbyn Calmly DISMANTLES Undercover Prank Needed to Restrain impenetrable noise of big data and their cognitive tractability is “one of big style Student who called On The Jeremy Kyle Violent Son | The the most important areas where political art could be situated today” him 'STUPID' Show - Saturday Jeremy Kyle Show Night Takeaway (Srnicek, 2012, 4), and - we add - a key area of design intervention. Moreover, faced with vast, unintelligible systems designed to capture 16 Jan 2019 10 Dec 2018 20 Feb 2014 14 Aug 2015 230,381 1,273,737 8,015,446 1,294,164 attention and orient choice for profit, the “inability to cognitively Elliot Etherington 50 Patriots Ant & Dec's Saturday Night Takeaway The Jeremy Kyle Show map the gears and contours of the world system is as debilitating for political action as being unable mentally to map a city would prove for a city dweller” (Toscano and Kinkle, 2015). How is any form of reclaiming to take place if the terrain is unknown? Ben: one of Jeremy "YOU'RE A "How Many More BBC Question Time What practical counter-strategies are possible? Kyle's most LIAR"German Referendums Do You 17/01/2019 DERBY disliked guest. Reporter INSULTS Want?" Rees-Mogg Nigel Farage! DISMANTLES Giorgio Agamben’s notion of apparatus is illuminating here as Watch His Response Remainer “anything that has in some way the capacity to capture, orient, deter- mine, intercept, model, control, or secure the gestures, behaviours, 22 Jul 2010 28 Dec 2018 27 Dec 2018 17 Jan 2019 opinions, or discourses of living beings” (Agamben, 2009, 14) with 4,476,928 157,979 153,184 112,973 a resulting dispersion and granularization of the self (Deleuze’s Phil Townsend ELITE PLUS 50 Patriots Kayn dividual, again). Only the restitution to common use of what has been captured can succeed as counter-apparatus interventions. Fox News Sunday Why did Mueller’s Hannity: Pelosi Democrat who This is what our proposals attempt to do. They use diagrams to map with Chris Wallace office choose to feigns moral opposed Pelosi for 1/20/19 - Chris speak out about outrage speaker is blocked the contours of the capture apparatus and crucially to offer some raw Wallace Fox News BuzzFeed report? from a seat on the glimpses of their inside. They use diagrams as tools that by display- January 20, 2019 judiciary ing how choice is designed as predictable patterning aim to be both committee practical and speculative. They are practical because they visualize 20 Jan 2019 20 Jan 2019 17 Jan 2019 20 Jan 2019 data-driven choice design by framing it through the challenges of data 36,831 46,732 286,428 189,099 positivism. They are speculative because they offer traces through Fox News Live Now Fox News Fox News Fox News which some partial and situated knowledge can be constructed, staying clear of interpretation and representation. Against the positivist claims to complete objectivity made by data visualisation and smuggled into Conan Furloughs so many of its tools, and unlike tracings (always reproductions with a Non-Essential likeness to their subject), traces offer opportunities for what historian Staffers Carlo Ginzburg calls ‘conjectural knowledge’ (Ginzburg 1980) - a mode of knowledge that is qualitative, contingent and incomplete, akin to div- ination (chance-led inquiry into the unknown) or the conjectures of the 2 Oct 2013 tracker (sensing an animal/event that cannot be directly experienced). 13,604,430 Team Coco #rtd2019 #researchthroughdesign #delft #rotterdam 12 13
Conclusion References We think that our programme of work is important for design research Agamben, G. (2009) What is an apparatus? Stanford, Stanford University Press for key reasons. Diagrams precede the instalment of a technology. For a technology to be possible, in other words, its materiality/ Albright, J. (2017) ‘ FakeTube: AI-Generated News on YouTube’ https://medium.com/@ machinery/toolkit have to be ‘selected’ by a diagram. By extracting d1gi/faketube-ai-generated-news-on-youtube-233ad46849f9 Frictions and Shifts in RTD the diagram expressed by current technologies of choice design we might develop counter diagrams that situate and dislodge its Ananny, M. and Crawford, K. (2016) Seeing without knowing: Limitations of the transparency mechanisms of preferences and recommendations. To do so, an un- ideal and its application to algorithmic accountability. New Media & Society. 1–17 derstanding of diagramming is essential especially for designers who do not design the technologies that they use. Arnold, S. (2016) Netflix and the Myth of Choice/Participation/Autonomy, in McDonald, K. and Smith-Rowsey, D. (eds), The Netflix Effect: Technology and Entertainment in the 21st Still, these initial and ongoing experiments are attempts at building re- Century, Bloomsbury lations, not revelations. In demonstrating the project’s feasibility, they provide an understanding of the tools widely used in the prior art on Boehnert, J (2018) Design, Ecology, Politics: Towards the Ecocene. London, Bloomsbury network mapping. They highlight how tools such as Gephi have many limitations for our purposes, starting with built-in assumptions that Boehnert, J. (2016) Data Visualisation Does Political Things. DRS2016 make tracing, rather than mapping, their default mode of operation. https://drs2016.squarespace.com/387 Importantly, they emphasize the challenge of using data to critique data systems, thus offering, we hope, a constructive reflection to an issue Bounegru, L. et al. (2017) ‘Narrating Networks. Exploring the affordances of networks that is central to doing research through design, that can be productive- as storytelling devices in journalism’, Digital Journalism, 5(6), pp. 699–730. doi: ly addressed through the theoretical framing of the predictive regime 10.1080/21670811.2016.1186497. of algorithmic governmentality and the preemptive politics it enforces. Bridle, J. (2017) Something is wrong on the internet. Medium https://medium.com/@ On this basis, some questions remain: how do we visually acknowledge jamesbridle/something-is-wrong-on-the-internet-c39c471271d2 the limitations of our system, conveying its partiality and incomplete- ness? How do we communicate the indeterminacies and uncertainties Burrell, J. (2016) How the machine ‘thinks’: Understanding opacity in machine learning through meaningful practice and knowledge production? Which other algorithms. Big Data and Society. 3,1 1-12 ‘architectures of choice’ might be diagrammed with this approach? Chaslot, G., Montcoudiol, A., Ledésert, S., Wielonsky, N., Bardolle, F. and Grac, M. (2016) AlgoTransparency [Online]. Available at http://algotransparency.org/ (Accessed 3 October 2018). Cheney-Lippold, J. (2011) A New Algorithmic Identity: Soft Biopolitics and the Modulation of Control. Theory, Culture & Society, 28, 6. 164–18 Acknowledgements Corner, J. (1999) The agency of mapping: speculation, critique and invention In Mappings, The project Architectures of Choice Denis Cosgrove ed. London, Reaktion 213-252 Vol.1 Youtube was intially com- missioned for Everything Happens Covington, P., Adams, J. and Sargin, E. (2016) Deep Neural Networks for YouTube Recom- So Much, an exhibition at the mendations. Proceedings of the 10th ACM Conference on Recommender Systems, RecSys London College of Communi- Boston, MA, ACM, 191–198 cation (LCC) for London Design Festival 2018. It was a collabo- Deleuze, G. (1995) Negotiations. New York, Columbia University Press ration between Supra Systems Studio at LCC and BBC R&D. Deleuze, G. and Guattari, F. (1988). A Thousand Plateaus. Capitalism and Schizophrenia. London, Continuum David Benqué’s doctoral research at the Royal College of Art is sup- Drucker, J. (2014) Graphesis. Visual Forms of Knowledge Production. Harvard University ported by a scholarship from Mi- Press crosoft Research Cambrdge (UK). Drucker, J. (2011) ‘Humanities Approaches to Graphical Display’, DHQ Digital Humanities Quarterly, vol. 5, no. 1 [Online]. Available at http://www.digitalhumanities.org/dhq/ vol/5/1/000091/000091.html. Ginzburg, C. (1980). ‘Morelli, Freud and Sherlock Holmes: Clues and scientific method’. History Workshop, 9. Oxford, Oxford University Press. pp. 5‐36 Hansen, Mark B.N. (2014) The Operational Present of Sensibility. The Nordic Journal of Aesthetics 24, 47. 38-53 #rtd2019 #researchthroughdesign #delft #rotterdam 14 15
Jiang, L., Miao, Y., Yang, Y., Lan, Z. and Hauptmann, A. G. (2014) Viral Video Style: A Closer Look at Viral Videos on YouTube’, Glasgow ACM Press, 193–200 Kitchin, R. (2014) Big Data, new epistemologies and paradigm shifts. Big Data & Society. 1–12 http://journals.sagepub.com/doi/10.1177/2053951714528481 Frictions and Shifts in RTD Lewis, R. (2018) Alternative Influence: Broadcasting the Reactionary Right on YouTube, Media Manipulation Initiative, Data & Society. https://datasociety.net/output/alternative-in- fluence/. Lewis, P. and McCormick, E. (2018) ‘How an ex-YouTube insider investigated its secret algorithm’, The Guardian, 2nd February. https://www.theguardian.com/technology/2018/ feb/02/youtube-algorithm-election-clinton-trump-guillaume-chaslot Marenko, B. and Benque, D. (2018) ‘Speculative diagrams: plotting to reclaim algorithmic prediction.’, in Art, Materiality and Representation. British Museum, London: Royal Anthro- pological Institute. Available at: https://theairpump.davidbenque.com/speculative-dia- grams-plotting-to-reclaim-algorithmic-prediction/. O’Sullivan, S. (2016) On the diagram (and a practice of diagrammatics), in Situational Diagram, eds. Karin Schneider and Begum Yasar, New York: Dominique Lévy, 13-25 Parisi, L. (2019) The alien subject of AI. Subjectivity. 10.1057 https://doi.org/10.1057/s41286-018-00064-3 Rouvroy, A. (2016) Algorithmic Governmentality: Radicalization and Immune Strategy of Capitalism and Neoliberalism? La Deleuziana 3: “Life and Number”, 30-36 http://www.ladeleuziana.org/wp-content/uploads/2016/12/Rouvroy2eng.pdf Shapiro, A. (2015) ‘The Next Big Thing In Design? Less Choice’ Fastcodesign. https://www. fastcodesign.com/3045039/the-next-big-thing-in-design-fewer-choices. Srnicek, N. (2012) ‘Navigating Neoliberalism: Political Aesthetics in an Age of Crisis’, The Matter of Contradiction: Ungrounding the Object https://www.academia.edu/1925994/ Navigating_Neoliberalism_Political_Aesthetics_in_an_Age_of_Crisis. Toscano, A. and Kinkle, J. (2015) Cartographies of the absolute, Winchester & Washington, Zero Books #rtd2019 #researchthroughdesign #delft #rotterdam 16
You can also read
