Although first alerted to Scarlett coffee in Angel by Double Skinny Macchiato last summer, we managed to visit during the one week of their summer holiday (and so we revisited Katsute100 around the corner instead). Nonetheless, it remained on the list and a few weeks ago we turned up for a mid-afternoon coffee at this inconspicuous looking venue on a side street just around the corner from Angel tube.
The roaster at the back of the cafe forms an immediate impression. With the large, communal table at the front of the cafe, backed by stairs leading up to the roaster, this is a place where coffee is taken seriously. The counter (on the left as you enter) offered a range of cakes and edibles but having recently come from lunch in Chapel Market, we passed on this on this occasion. Above the counter there were about 5 lights hanging down forming what looked like a giant Newton’s Cradle. Just too high for me to reach unfortunately.
I enjoyed my long black as I started to take in the surroundings of this cafe. Various people and regulars came and went, suggesting that this is a friendly local haunt for many. Noticing the number of different roasted coffee beans for retail, it was clear that this is a venue that you could return to for a different coffee experience each time. Each time exploring an aspect of the flavour of the coffee and building on the experience of coffee tasting that you have enjoyed before. It is definitely on the list for a repeat visit.
Above our heads, the lights were framed by the outline of a cube. Fantastic for optical illusions, these cubes offer us an opportunity to think about how we perceive depth and direction; how our eyes work and perhaps, more fundamentally, what it even means to see an object (as with Berkeley’s “New Theory of Vision”). Then, while looking through the menu, it became clear that here too there was an optical illusion of sorts. For the price list was not written on the board so much as cut out of it (see the photo below). The price you could read off the menu was, in some sense, precisely the information that was not actually on the board. Our brain makes patterns of that which we don’t see and, together with our assumptions about what should be there, we form an idea of the price we have to pay.
It is a similar thing with many algorithms in use around us now. Such tools can be immensely helpful, offering us suggestions for coffees we may like to try (based on our buying habit) or routes that we may like to take to get us to our destination. And yet, are there problems hidden in the assumptions that some of these algorithms make? What information are we getting based on elements in the programme that we do not see?
In her excellent book “Weapons of Math Destruction”, Cathy O’Neil explores some of the more dangerous ways that our biases and assumptions (particularly those that we don’t see in ourselves) can impact the results of algorithms that have been written to optimise processes from the sorting of job applications to determining the length of time a given convicted criminal will serve for an offence. In an example relevant for cafes, O’Neil related an example of how Starbucks had used an algorithm to determine which baristas and managers should work which hours, including who should close the shop at night and who should open it in the morning.
The algorithm was programmed to calculate the most efficient use of the cafe’s time and money, specifically prioritising the profit that the company made. One measure of this was “revenue per employee hour”. This had the consequence that staff members were frequently in a position where they were told that they had to do both the (late night) closing and (early morning) opening of the shop and were given very few days notice of this expectation. Clearly this impacted the lives of their staff and affected their ability to arrange child care, support themselves through further education and other consequences. Eventually Starbucks was forced to amend this algorithm but change comes hard: how do you ask a computer to measure “fairness” to an employee (a subjective term) when you can use revenue per employee hour which is measurable, quantifiable and therefore ‘accurate’?
Perhaps you think that the link back to Scarlett here is obvious: That if you choose to drink your coffee in friendly neighbourhood cafes where cafe owners and baristas work to patterns formed by encounter rather than algorithm it would be better than a place which is run assuming all workers are cogs in a profit machine? Perhaps. But the link back to Scarlett in my mind is not that at all.
If you look at the front of Scarlett, or its webpage, and assume that the pink bird is a funny looking flamingo, you may make a series of assumptions about what you think the cafe will be like and why the owners have a bird on their front door. If you found out that the bird was actually a Scarlet Ibis and associated with the coffee growing regions of South America, your ideas about the cafe and the owners may be different. For a general customer, looking for somewhere to enjoy a great coffee, perhaps these assumptions and ideas do not matter so much. But if we are ever in a position to feed our assumptions into an algorithm, these hidden (to our own conscious) assumptions could matter a great deal.
Scarlett is at 30 Duncan Street, N1 8BW
“Weapons of Math Destruction – how big data increases inequality and threatens democracy” by Cathy O’Neil, Penguin Books, 2016
