underconstrained
conversation and inspiration
You’re given a blank canvas and there are no rules, no expectations, and no pressure. Just three individuals with the task to create something.
So where do we start?
Well in our case it started with a lot of…well… chatting. Talking about random things: current events, personal interests, Pennsylvania prisons. It didn’t feel like anything was about to coalesce or that there was any major push towards clarity. It just felt like people getting to know each other. Which was casual, but important because we needed to form an initial understanding what each of us could bring to the table. This included skills, things we’ve worked on, and possible visions for the project which is what Ying (our artist) shifted our attention to next.
Themes
Ying’s prior work has mostly been contemporary commentary about Chinese censorship and internet culture behind China’s infamous great firewall. With China-US relations being a constant source of gaffes and headaches there is certainly a lot of current material to work with; however, amid all the trade war news there is one development that is of particular interest to us: implementation of the Chinese Social Credit System.
In a nutshell, China’s major plan is to create a comprehensive credit system where it can use any and all sources of information about individuals to assign them a score. This score then dictates what privileges and restrictions that could be imposed. For example, a citizen with a low score could be restricted from access to air travel, denied certain job prospects, or even be subjected to public shaming.
If this sounds like a familiar dystopia, you probably have seen or heard about the Black Mirror episode Nosedive. In their version of a social credit system people are given the ability to rate their fellow neighbors, colleagues, or friends based on how positive or negative they view an interaction — akin to an Uber user rating a driver and vice versa. Even if you haven’t seen this episode you can imagine how restrictive — and potentially unfair — such a system could be; but at least there is some level of transparency afforded to the user. They can immediately see who is affecting their score and, based on their interaction, can understand why someone rated them as such. This is the key difference between this fictional depiction and China’s proposed system: Transparency.
This concept of China’s methodology being shrouded in ambiguity became our first source of inspiration which resulted in us experimenting with abstract data visualization
In China’s system you are rated by the Government based on data that they have aggregated across various sources, but if you try to find a detailed explanation of what exact data is being used and how it’s being weighted it is unclear. This concept of China’s methodology being shrouded in ambiguity became our first source of inspiration which resulted in us experimenting with abstract data visualization.
Ideas and Initial Visualizations
The purpose of visualizing data is generally to communicate a clear message to the viewer. In this case, our motivation was not to communicate an obvious message. Instead, we wanted to use data to illustrate a concept in an abstract and layered way which in some sense emulated the ambiguity of the social credit system. In order to attempt this we looked at various different data sets and visualization tools to get a sense of what might be visually appealing — some of which is illustrated above.
This is great and all, but there is one fundamental issue to consider: if we pursue this further, how will we get any relevant data? The social credit system itself is a mystery and it would more than likely close to impossible to get access to relevant public data sets. This prompted us to discuss the idea of convincing data generation which — as one thing leads to another — steered our conversation towards current trends in machine learning.
Moving Forward
Similar to how our initial conversations had begun before, we transitioned seamlessly into an exchange of knowledge and ideas. My partner Michael and I helped get Ying up to speed about the current applications and capabilities of today’s machine learning algorithms which opened new doors and opportunities for experimentation. Techniques like Generative Adversarial Networks (GANs) and Neural Style Transfer quickly set our sights beyond simply creating visualizations of fake data. We were now throwing out ideas like: “What if we generated fake people that was seeded by a fake credit score…What if we made a bot that learns to judge people’s chat correspondences and patronize them accordingly…What if we try to mix Brad Pitt’s face with China’s leader Xi Jinping?
While we probably won’t be able to explore all the what-ifs we posited over the past weeks, we have filled our blank canvas with a variety of ideas, thoughts, and visuals from which we can explore and develop.
Next we will be delving deeper into more nuanced thematic content (see Ying’s post to get an initial sense) and the machine learning techniques we used to explore this content.
