KDnuggets Home » News » 2015 » May » Opinions, Interviews, Reports » Interview: Haile Owusu, Mashable on Surviving Imprecision in Digital Media Analytics ( 15:n14 )

Interview: Haile Owusu, Mashable on Surviving Imprecision in Digital Media Analytics


We discuss the challenges in tracking social media sharing, advice, important trends, and more.



haile-owusuHaile Owusu is Chief Data Scientist at Mashable where his main responsibility is the development and refinement of the company's proprietary Velocity technology, which predicts and tracks the viral life-cycle of digital media content.

Prior to joining Mashable, Haile led all research efforts for SocialFlow, one of the leading social media optimization platforms for brands and publishers. Haile specializes in statistical learning as applied to predictive analytics and has a background in theoretical physics, including a Ph.D from Rutgers University, a Masters of Science from King's College, University of London and a B.A. from Yale University.

First part of interview

Here is second and last part of my interview with him:

Anmol Rajpurohit: Q6. What are the most underrated challenges in tracking the consumption and sharing of digital media?

Haile Owusu: Though I think it’s acknowledged that predicting broad scale social behavior is very difficult, our task is particularly difficult because we are confined to very bare data. Most studies of social diffusion operate with more granular information than we have. Understanding how the flu spreads or how email chains propagate in a known network is difficult even when you have intimate knowledge of how the people involved interact or, maybe, broader topological features of that network. Because we and the social networks we study very much respect the privacy of their users, we’re largely blind to such network topology. This makes a useful prediction that much harder.
social-media-challenges
AR: Q7. What is the best advice you have got in your career?

HO: It wasn't an explicit piece of advice, but a question: what is your risk profile? At the time I was job hunting, not long out of academia and it was asked of me by a senior data scientist. riskThe question was extremely clarifying because it crystallized for me how I want to spend my time and leverage my expertise and in what sort of environment. As it happens I prefer working on unsolved problems and with colleagues who can iterate aggressively towards a solution. Velocity has been a tremendous joy to work on and Mashable, along with my exceptional colleagues, very much provide that environment.

AR: Q8. Which of the current trends in Data Science are of great interest to you? Why?

HO: I am particularly fascinated by the cascade of results coming from computer vision. Why? ...well I think I have the same romantic fascination most people do with the idea that a computer can ‘see’ an image and ‘know’ what it contains. Given how intimately bound human vision is to how we understand our environment, a computer’s ability to mimic this is extremely computer-visioncompelling. More concretely though, the most compelling advances in computer vision essentially generate descriptive, human readable sentences out of images. Given this linguistic analogy, working on Velocity we imagine that the extent to which an image subject matches the interests of user’s social network affects its shareability. Beyond that though, there’s a much harder to characterize component of imagery that operates visually, much like the prosodic elements of language operate linguistically. Tack a compelling beat onto a series of rhyming couplets and you can turn incoherent words into lyrics people will share for a generation. I wonder what the visual analogue of that is and how best to bring that to bear on the content we see shared in Velocity.

AR: Q9. What key qualities do you look for when interviewing for Data Science related positions on your team? interdisciplinary


HO: It’s hard to generalize because I think an effective data science team working on interesting unsolved problems has to be quite interdisciplinary, with representation from my own discipline of physics, engineering, mathematics, biology, the social sciences. The only unifying attribute I absolutely need these representatives to have is demonstrably indefatigable curiosity.


AR: Q10. What do you like to do when you are not working?

HO: Not working? What is this concept you speak of…?

I enjoy interactive theater, getting better at agile street photography, the occasional voice over gig. And booze.

Related: