Chan Zuckerberg Meta: designing a meaningful AI tool for life science researchers

In 2016 I joined Meta – a small, but scrappy Toronto-based startup working on a product that helps researchers understand what is happening globally in science using AI and machine learning. When I started, the free application for ingesting research papers was changing every day. Our free application for researchers provided good signal, but we needed to commercialize the knowledge graph. We quickly tried our best to figure out what the large-scale applications of product could be and how they should feel.

My first task as the only designer on the team: help find product-market fit for our technology, work with researchers on designing a suite of products based on their needs, and in the process craft a new brand identity and a design system for the newly-named product.

Understanding the research cycle

We know that science has a problem. With over 4,000+ new papers published each day in biomedicine alone, scientists are struggling to keep up with relevant research.

To build a suite of successful commercial knowledge-based products, we first needed to understand:

Who are our potential users and what is their research cycle? What are the bottlenecks?
How can our unique insights bring value to them and how do they fit into their workflow?
How can we predict if scientific information is promising as it’s published?
What does our data look and feel like?
Is there a commercial market for products that address these problems?

In addition to traditional user research, we had the advantage of drawing on learnings from experiments carried out in our data science lab. In an attempt to bring clarity to our understanding of user workflows, we started visualizing key learnings.

Looking at related concepts in a variety of fields allowed us to gain a deeper understanding into how knowledge builds.

Visualizing the Meta Knowledge Graph resulted in a beautiful representation of science – a breathing, living map of interconnected concepts and collaborators.

Mapping the social nature of emergence allowed us to gain an understanding of how emergence can be predicted with large data sets.

Looking into the paper submission process, we found statistically-significant factors that determine what make a certain manuscript successful in being published. The visualization above shows chances of success for a manuscript going through a journal submission process.

Through a process of rapid visualization, whiteboarding and prototyping using D3.js we identified a few unique visualizations that brought our data to life and surfaced relevant insights.

Building commercial applications

Based on the outcomes of our research and experimentation, we designed and built several commercial desktop and mobile products on top of the existing knowledge graph, including:

Bibliometric Intelligence

A tool that enables editors with unprecedented insight through machine learning.

Horizon Scanning

A tool for scanning the scientific, technical and medical horizon for signs of future emergence.

Defining our design methodology

We started by identifying the following principles to guide our work.

Support scientific method and accelerate scientific process
Understand the user demands and our abilities to bring simplicity and user-centric design to scientific community. Design tools that make a big difference and accelerate paper discovery.
Enable discovery and “eureka” moments
Empower researchers to go deeper to surface new connections and insight. Act as a catalyst for magic moments of discovery.
Encourage collaboration
Design experiences that fill the collaborative void and encourage creative potential in global scientific community, mentorship, and professional growth.
Experiment, refine and iterate
Embrace and learn from all experiments, trial and error, and failure. Break convention, explore all potential solutions and ignore bias. Grow by learning.

In approaching the interaction design for both products we focused on reducing friction in their research workflow.

Supporting interactions focused on ease-of-use, while surfacing key information relevant to the researcher.

A visual language was created around some of the new concepts we were introducing to the world of science UI.

Primary colour scheme was inspired by microscopy staining. Contrast in colours for visualization was a key consideration.

Refining the brand

As we were busy building the commercial products, we had to perform a wide range of market tests of our messaging and look and feel. This involved an endless stream of of email campaigns, A/B tests, optimizing landing pages for conversions, refining our target demographic, building presentations, videos and other collateral.

A series of videos were produced to educate users on the power of Meta.

Educational collateral and white papers helped users gain a deeper understanding of product potential and put us in touch with the most qualified and interested audience.

As we were busy heads-down working on the product and acquiring customers for the commercial applications our work started receiving press coverage by outlets like TechVibes, Globe & Mail, betakit, CB Insights and others.

“Most scientific breakthroughs have been preceded by the invention of new tools that help us see and experiment in new ways.”

Mark Zuckerberg
Co-founder, Chan Zuckerberg Initiative

On January 23, 2017 it was announced that the Chan Zuckerberg Initiative was acquiring Meta, empowering us to carry out our mission on a brand new scale. Receiving this news with our small team was one of the most rewarding experiences of my life.

Fast-forward to today…

After the acquisition, our focus rapidly shifted towards a new, reimagined vision for Meta. One that empowers researchers and accelerates the progress of science at an unprecedented scale. We were able to rapidly onboard an amazing design team that helped us accelerate our progress in launching the new version.