In late July, myself and my colleague Marko Satek were invited to attend the 2018 Trusted Media Summit, an event organised by Google in Singapore which brought together journalists from different countries of the Asia-Pacific area and technology companies like Storyful.
The goal of the event was to discuss tools that could help journalists in the battle against misinformation. We were fortunate enough to talk to many people and understand the difficulties they face in their investigative work. They emphasised how severe this problem is in some parts of the world, where misinformation can have serious consequences on people’s lives.
Technology is not meant to be used as a replacement for human judgement, considering the complexity behind different cases and the importance of context when verifying pieces of information. But technology can help to increase workflow speed, and identify patterns in high volumes of data. That’s what computers are good at, right?
Following two busy days of presentations and panel discussions, Google had scheduled a three-day design sprint with no pre-defined objectives in mind. The sprint team was made up of a small number of people with different backgrounds: two journalists, three UX experts, five engineers, and three students from the University of Singapore.
The journalists demonstrated all aspects of the investigative process, presenting some real use cases, while we annotated what could be potentially improved. This task was called “How Might We?”
We then presented our notes, grouped them into clusters and voted on relevance. This helped to identify two areas of improvement:
- the task of preparing a picture for an image search — cropping, rotating and flipping — is very time-consuming for journalists. There was a need to automate this process.
- the results of an image search were ordered by recency, but journalists were more interested in older results.
The second part of the design sprint was about sketching a few ideas and solutions, starting with an exercise called “The Crazy 8.” It consisted of drawing eight quick solutions to the problem identified in the first step. After voting we all started to have clear in mind the app we wanted to build. In fact, the following exercise was intended to draw a complete user journey and then vote on the best solution.
The next phase was surprisingly quick: we gathered around a board and quickly designed the main components of the app.
We suggested Angular v6 for the client, which seemed the perfect framework for the goal, especially because it comes with the angular material package, which was welcomed by the UX experts. We also decided to follow a mobile-first approach as journalists expressed the need for this app to work mainly on mobile.
We then identified three main services with individual responsibilities:
- image processing (crop, flip, rotate)
- image search
- text extraction and translation
In particular, the last feature was very important for journalists, especially in countries like India where the variety of languages can be a challenge. We leveraged the power of Machine Learning via a pre-trained model provided by the Google Cloud Vision API. Timothy, Tong and Chester, students from the University of Singapore, took ownership of these services and resolved to build them in Python.
The last piece was an intermediate layer with the following responsibilities:
- handle requests from the client app
- query the image-processing service, the search service and the text extraction service
- format the results and present them to the client
For this part we decided to build a Rails API app.
At the same time the UX team was preparing the mock-ups and giving us directions to make the app intuitive and user-friendly.
We were taken aback at how everyone seemed to be on the same page in such a short time and we think the open-ended nature of the design sprint facilitated this. It was late afternoon when everyone grabbed their laptop, a good cup of coffee and started typing.
Believe it or not, after one-and-a-half days, and a lot of coffee, we had our prototype built in all designed aspects, nicely styled and deployed to Google Cloud. We were ready for the demo.
To an audience of around 40 assembled journalists and technologists, the result of our work was demonstrated by Shammas Oliyath, a journalist from India. He uploaded two images from a project he had been working on, cropped the portion he was interested in and performed a reverse image search to find earlier versions. The app then extracted the text from the image so he could translate it into multiple languages.
The demo went very smoothly and the audience appreciated how much we’d built in such a short time. Many even asked for access to the prototype. This is probably the most rewarding thing a software engineer can experience in their career, the positive feedback of solving a problem.
On our flight back to Ireland, Marko and I chatted about the experience in Singapore. We both agreed that a design sprint is a very useful framework in helping teams to find solutions to complex issues.
The process is also useful to ensure people are on the same page.
In Storyful we have attended other hackathons, but the big difference with design sprint is that the goal is not the prototype we want to build, but discovering the problem we want to solve.
Thanks to Irene Jay Liu and all the engineers at Google for organising the 2018 Trusted Media Summit and facilitating the design sprint.