Viziflu / COVIZ

 

Data Viz, Uncertainty, Forecasting, Influenza, COVID-19, User Study

Viziflu is a browser-based visualization tool that helps analysts compare infectious disease forecasts, understand the uncertainty in predictions, and determine when seasonal influenza is likely to peak.

Developed in collaboration with the CDC, Viziflu displays crowd-sourced forecasts submitted to the FluSight forecasting contest. The tool makes it easy to compare forecasts produced by different models and understand each model’s predicted probability that flu will peak in a given week of the flu season. In April 2020 we adapted Viziflu to display peak forecasts for COVID-19, creating a new tool we called COVIZ

Viziflu is an IQT Labs project, developed in collaboration with G. Sieniawski & D. George, based on input from the Centers for Disease Control & Prevention. Illustrations & animation by DarkCharm. For more info, check out: What flu forecasting taught us about visualizing uncertainty. 

VeriCAT

 

 

 

Interface Design, Machine Translation, Quality Estimation, User Study

Machine Translation is good, but not perfect. How do you know when to trust it? 

VeriCAT, short for Verification of Computer-Assisted Translation, provides contextual information about the quality of machine-translated text. The tool uses a Quality Estimation (QE) model to predict the translation quality of Russian sentences that have been translated into English with fairseq. VeriCAT’s simple interface makes quality scores easily accessible, to help people determine when to trust the translation. 

VeriCAT is an IQT Labs project, developed in collaboration with N. Lopatina, J.Pope & A. Mosca. Our paper How Good is your Machine Translation? Design and Evaluation of a Quality Estimation System is currently under review.


What do people really want when they say they want Explainable AI?

 

User Research, Explainable AI, Qualitative Research

In 2019, I interviewed 40 people and conducted 2 focus groups to explore what Explainable AI (XAI) means to different stakeholders in tech & government. As it turns out, this buzz word means *very* different things to different people.

I summarized my findings in   this short paper and published excerpts from several interviews on High Stakes Design , including this post with Outlier.ai CEO Sean Byrnes and this, with Fiddler Labs CEO Krishna Gade. For more info, see: 10 things I’ve learned about Explainable AI  and Explainable AI: what is it and who cares?


CRviz

 

Data Viz, Networks, Cyber security

Do you know what’s on your network? Modern computer networks house tens of thousands of devices, many of which are constantly changing. This makes it extremely difficult to keep track of what devices are on a network at any given time. 

CRviz is a browser-based visualization tool that — when used with Poseidon, an open source network analysis tool — shows you what’s on your network. With CRviz, you can organize devices into groups and sub-groups based on attributes you choose. And because CRviz uses a visualization technique called “circle packing” you can finally say goodbye to  hairball visualizations .  

CRviz is an IQT Labs + CyberReboot project, developed in collaboration with C. Lewis, G. Shipley, R. Ashley, A. Chang, H. Tambala and and Taivara. Illustrations and animation by DarkCharm. For more info, check out: Scalable design for network visualization.

dataviz.cafe 

Data Viz, Open Source

Dataviz.cafe is searchable catalog of 700+ open source visualization tools, tagged by programming language, data type and topic. We built this site to help people find the right visualization tool for their data. If you know of a tool we’ve missed, follow  these instructions to submit it!  

Dataviz.cafe is an IQT Labs project, developed in collaboration with G. Sieniawski. 


 

Wandering Dots

Data Viz, Uncertainty, Mapping, Human Geography

If you’re not 100% certain where something happened, how do you show error bars on a map? 

We didn’t know either, so we designed a visualization technique called “Wandering Dots,” which uses animation to show uncertainty in geolocated data points. We used this new technique to visualize the locations of conflict events in South Sudan, based on data from the Armed Conflict Location and Event Data Project (ACLED).

This work was part of IQT’s 2017 Mission Challenge. It was developed in collaboration with N. Miller, R. Pappu, K. Pernia, L. Porter, and Bocoup. For more info, see: Visualizing Uncertainty in Human Geography Data.


 

NetSight / NetSite

Data Viz, Networks, Routing

From 2011-2014 I managed the design & development of several visualization tools to help communications network researchers and engineers improve the performance of MANETs (Mobile Ad Hoc Networks). This was my first introduction to the difficulties of visualizing complex graphs and the infamous “hairball problem” (i.e. when you try to visualize graph data and end up with something that looks like a hairball). 

Most of this work was not published, but there are a few exceptions here   and here.

 


 

The Office for Unsolicited Architecture 

 

Before data viz & UX design, machine learning and software, I was a founding member of the Office for Unsolicited Architecture (OUA).

****

As it is practiced today, architecture is too often a responsive profession. Architects take direction from their clients, providing solutions to a set of problems and desires described in a project brief. While this sort of responsive architecture is, admittedly, the most pragmatic way of working (since financial remunerations are secured in advance), it affords little agency to the architect. Clients alone determine architecture’s agenda.

OUA proposed an alternative: pro-actively seek out new territories for design intervention, addressing pressing social needs while taking advantage of emerging opportunities for architecture. We were motivated by a desire to fight passivity, self-doubt and perceived (or real) marginalization of architects, while hunting for an architecture that is interesting, innovative, subversive, creative, transgressive, reflective, sustainable, attainable, idealistic, profitable and above all, forever unsolicited.