The Potential of Digital Games to Engage Citizens with Datasets
We started this project staring down a massive collection of text data that the European Commission had handed over to us, full of their written science reports that have accumulated over the years. The reports suffer from anonymity to the average citizen, though they are open and transparent, publicly funded and gathered for public benefit. The breadth of topics is seemingly impossible to relate back to our everyday issues and the wicked problems that fill our newsrooms, daily concerns, and dinner conversations.
How much anonymous and unstructured data is out there for our potential benefit, but we don’t perceive it as attainable or insightful? How can we contextualize this data in an emotional or polemic way, connect it to the issues we care about, and uncover more insights through participation?
These are the questions we asked ourselves when The Joint Research Centre with the European Commission prompted us to make an interactive experience for people to play with the massive knowledge database, where scientists measure anything from emissions in Germany to soil quality in Italy. How can we playfully entice people to interact with such a large (and maybe a bit boring) database?
The Disjointed Research Machine is an attempt to shift the original reason d’être of this database. We designed a playful experience infused by machine intelligence that unapologetically disjoints the data into pieces for more creative exploration and combination. Our approach to what we are calling disjointed research is about investigation and experimentation that embraces chaos, machine randomness, and human intuition.
Setting up a test for colleagues to play the game
Inspired by the irreverence of games like Cards Against Humanity® and Mad Libs®, and the intuitive rules of BananaGrams® and Word Jumble, we created a digital word game, where players set up their own provocative challenge, like ‘create a political slogan’ or ‘write a tabloid headline’, and respond to it using only phrases taken from the science database. In one minute, you must rearrange words and assemble your response to the challenge. In essence, players are challenged to contemplate a wicked problem and a polemic issue, that is interesting for them, but then address it with the language of the EU science reports.
As a one-player game, we still wanted the experience to be social. So though the game is played on your own phone, a large screen shows your progress, and the results of players before you. Responses to challenges are graded and compared so that you can see what other people have come up with so far.
The technical foundation of the game derives from research into semantic similarity in text data. A pre-trained machine learning model encodes human words into multidimensional vectors. This Word2Vec model gives us the ability to model a sentence as a vector in space, letting us measure the conceptual distance between sentences, ie. vectors closer together in space are closer together in semantic meaning. This allows us to search the database for sentences similar in meaning to the input, such as interpreting the statement “I fear climate change” and then outputting sentences from the database that have similar semantic meaning to this statement. This is also how the score is created, which is based on several criteria, such as if it appeals to the emotion you chose or uses words similar to those in your challenge statement. The final score is calculated from the semantic similarity of your answer to each of these criteria.
Addressing a dense dataset with an air of playfulness makes it approachable. Players might never have otherwise interacted with the EU science reports, though they are lawfully open and transparent, available for anyone to read on the web. Our game isn’t necessarily competitive in the traditional sense, though it does use gamification to draw people in, and play multiple times, compelled to achieve a higher score.
What if other public data sets could also foster interaction in a playful way? Can social games make connections between insights in data and the wicked problems that emotionally engage us?