Mark Simpkins and I took a very different approach to the Honda Hackday to the ones that we both take both at hack days and in our working lives. We decided to make some art together. Mark’s talked a bit about it here. We jokingly called the movement the Algorithmicists. Both of us have wanted to explore how you make things to provoke thought and invite inquiry in the observer and have no utility or purpose apart from that. Mark has for a while talked about an organisation of unknown shape called This is Our Algorithm and this was a chance for us to try it on, to see how it felt and many thanks to Rewired State and Honda for allowing us to.
Mark talks very well about some of the thoughts and principles:
“All three pieces dealt with algorithms, code and control. They were as much about some strange desire we have to codify up our lives into possibly complex but ultimately meaningless algorithms, trying to reduce complexity to a point beyond understandable simplicity to end at a nihilistic pointlessness.
Reality is complex, understanding reality is hard.”
The first piece and physically the largest piece we made was called A Dream Within A Dream. It is an exploration both into influence and algorithmic scoring and also a piece to provoke that thought in the viewer. It’s partly inspired by this fabulous tweet from Chris Sacca.
I have some fairly strong views about Klout. I really don’t like it in so many ways. I can sympathise with people who feel there need to be some form indicator about the reputation of an account on Twitter (real person vs bot) but I feel really rather sad about things like Klout in many ways. They dehumanise networks in my opinion. One use case I often hear is about looking at how companies can avoid very public confrontations with people by understanding who the influencers are and then ensuring that they’re helped/offered special treatment/pacified. This feels like an own goal for a quasi-meritocratic society at best and a re-enactment of the rise of the 1% at worst. Giving free things to influencers feels like giving gift bags to people at Awards Ceremonies.
I also have strong views about black box ranking systems or ones where the walls are so opaque that an everyday user cannot begin to see what is going on in the system. I can imagine which tweet it was that made Klout for a long while think I was influential about Hugh Grant. However it’s really quite an outlier in my stream, although I sent another today. I’m not an expert statistician, but as a former research scientist I find it worrying that a large amount of faith is put into a metric which clearly in some cases amplifies aberrant signal over noise and pretends it’s a relevance. I’d love to know how they run controls in their system. How they test it against randomness. How the look for and guard against false positives. How they compare against standard texts and against nonsense tweets algorithmically created out of common word tables.
Our way of probing all of this was to create a fake dreaminess index. To show the futility of algorithmic scoring and how you can create an index about anything against anything. To show through parody that value only exists in creating a truly realistic index if that is what you feel you must do. To show this we set out to make the most pointless one we could.
The first point in our plan was in calling it a dreaminess index as shown in the parody image above from our presentation. What are we referring to as dreaminess? The second point is to think about what a dream is. Like influence, the word dream, has so many meanings and nuances - from the dreams you have at night to short term and long term aspirations, both personal and for the wider world.
The mechanism by which we went about creating the index was a further probe. Indexes need to be compared to things. The only way to measure things algorithmically is against some form of ideal, be it a standard or a derived standard. We decided to create an index of dreaminess based on pure lexicographical similarity to Edgar Allen Poe’s “A Dream Within A Dream”. I could argue, as pointlessly and as vigorously as I liked if I wanted to, about why I think that poem is the very ideal of a piece of writing about a dream. Without statistics, and all of the things we as scientists take for granted such as reproducibility and controls it and any index or scoring table is pointless. We’d like to thank Dom Hodgson who suggested our +D index should be called Doubt. Inspired.
The way we created the index is deliberately meaningless too. The scoring table is based purely on word frequency analysis of the poem. Things you’d normally stopword out are deliberately left in. The poem is our ideal of dreaminess and we’re futilely comparing how similar all the words including “a” and “I” in your tweets are to the poem. It’s deliberately wrong and pointless. The number of times the indefinite article are present in the tweet are doubly amplified, it’s both a high scoring and common word. It has all the wrong forms of bias.
We invited people to send in their dreams to us, we set up an account called GiveUsYourDream. The wording was deliberate - we would become the possessors of your dreams. You gave them freely, we have them now. When not enough people opted in we simply did what all other similar indexes would do, we just hoovered in tweets from a Twitter search for “dream” without asking people to opt in. This again is a deliberate response and provocation, and in addition it ensures that every dream gets at least a score of 5 for having the word “dream” in it.
What we did next is to start making the physical side of the project, the visible part of the art work that people would see and engage with. We started writing down the dreams on blank pieces of paper chain. We wrote the tweet on the inside, so that it would be hidden partially from view. On the outside was the name of the person who had the dream and our +D score of their dream and, by inference, them. The reality, the sentiment, was barely visible when the chain was assembled; it was partially or largely obscured by the nature of the paper chain. The score and the name were dominant, just as it is on Klout. You’re boiled down to a score, the thing which you said is now a second class citizen.
Part way through writing down the scraped tweets I came across this one and it stopped me cold. It made me think that actually anyone who does anything algorithmic should have to, as part of what they do, physically write down some of the data. You get a new emotional connection to it. It becomes real. And personal. It’s clear that someone typed it, just as clearly as you are having to laborioiusly write it. I wanted in a way to share a picture of what we’d done with this person, but then I felt I may spook them out. Instead I’ll look at this from time to time as almost a cautionary tale to anyone who may think about sentiment or influence mining. People write things with hopes, fears, feelings - read and use those things with respect for that.
We then set out to assemble. We’d originally considered going the whole height of the Guardian’s floors at Kings Place but the physicality of writing and making the chain meant that we could only manage the height of one floor in the time allotted, even though it was an inspiring and moving sight. A chain of semi obscured dreams. Independent dreams now interlinked. Dreams inextricably held within other dreams. Dreams within dreams. The whole piece was designed to be utterly self referential, the text selected as the source of the scoring table related to the physicality of how the tweets were displayed and gave the piece its name. It was moving to see it installed, hung on transparent nylon thread, running through the void of the stairwell and gently touching the floor. Rain Ashford took this picture of it, the best I’ve seen, as it disappears up into the illuminated stairwell.
I miss it. It was something Mark and I talked about and schemed and thought about all weekend and it was there for a few hours and then it was gone (possibly to appear again we hope somewhere in Honda’s office in the UK). I do however have a feeling we’ll do more of these things. We spoke to many people afterwards who were really encouraging and said it had made them think about Klout and algorithmic scoring. We hope it makes you think about that to.
Footnote/Full disclosure: After my post bemoaning things on a weekend, yes, the Honda hackday over the weekend. I was paid to be there.
