PornHub.com : Terrible Data

-
Some years ago Pornhub.com was asked to gather and analyze data for BuzzFeed to answer a seemingly simple yet statistically involved question. The question was “Who watches more porn, the Reds (republicans) or the Blues (democrats) ?” Aside from the fact that this question was probably not the most efficient use of their time, BuzzFeed relied completely on the prowess of Pornhub’s statisticians and declared that it was evident from the analysis that the Democrats took the lead with a 13% difference. It so turned out that the data mining process flawed. This led to a lot of seemingly plausible insights which were later proved to be inaccurate.

Here are a few mistakes they did :
  • The first step was to obtain data about state-wise distribution of republicans versus democrats. Now you would think this can emerge from the data about the percentage of voters per state. But initially they assumed that just because there was a democratic victory (or a republican victory) in a state that all its porn-viewership must be counted towards the winning category. It was easy to see how this was flawed and this was corrected in time.
  • How can any company involved in the business of porn violate the basic right to voter anonymity for data gathering. It cannot ! Therefore, they collected data based on viewers without actually having an identity. This is the first step where they should have realized their mistake. The number of people watching porn in a state has got nothing to do with the number of people who go to vote. Therefore, trying to obtain any correlation in this regard would be a waste of effort.
  • The technology used in data gathering plays a key role in directly impacting the accuracy of data. It also indirectly helps reduce the effort required to clean the data. A good data scientist will always take into account uncertainty associated with data input, even if it was from the belly of the bleeding edge of technology. The data gathered by Pornhub.com was based on the IP address repository they had and they used a technique called geocoding which is known to give only a “likely” location because networks like VPNs provide a fake location. Even if we assume that very few people have VPNs, we cannot ignore the fact that a large population must know of the incognito feature in the browser that actively hinders the accuracy of this technique. Hence, what was most surprising is when the data was laid out, people of Kansas state turned out to have the most porn viewers. The reason was that when the geocoding fails to find a precise location, it returns a vague “middle of USA” location and that is where Kansas is situated geographically.
In conclusion, bad data practices can lead to an ineffective data analysis, most of which can be prevented by not letting data take place of common sense. It is important to remember in this regard that data analysis can aid an analysis but should not become the only point of reliance.

 In case you're wondering how old this was - it was 2014.

 References
  • https://source.opennews.org/en-US/learning/distrust-your-data/

Comments

Post a Comment

Popular posts from this blog

Why computers aren't amazingly smart

-
Artificial intelligence is the field of computer science with the goal of making computer more useful than they already are by making them receptive to their environment, giving them a novel way to store, process and expand their knowledge base and letting them make informed decisions based on the accumulated knowledge. As you can already see from the above definition that the field has a variety of sub-fields or places where one can contribute. Let's begin by talking about what motivates artificial intelligence. How can the task of making decisions in machines ( which is mostly inspired by how humans think and act ) be understood as a process ? The best way to draw a parallel is to keep in mind that what the machine is essentially trying to do is act like humans albeit without a perfect resemblance. The state of art thinking machines are not even as good as a teenager. In fact, most would not be able to do simple tasks like differentiating a girl from a boy as fast and as accur...
Read more

Making Machines : Introduction

-
Machines are now ubiquitous but once they were just thoughts in a crazy man's mind. Well actually, many men but for our understanding there were three people who set the ball rolling. 1. Alan Turing 2. Alonzo Church 3. Kurt Godel All these men were, in their own ways, trying to figure out a mathematical model for a machine. For all those readers who lost me here should think of a mathematical model as the way of describing anything around you with pen and paper. It's almost like taking notes but mathematics is the art of taking notes about your notes and doing all that till you make sense of what is possible by twisting what you know in ways you couldn't have directly imagined. Math is just a language and it's not necessary for all of us to think in that language. My blog here is another effort among many to explain what "computation" means and how is it different and similar to the word "computer" we use every day now. Think about what yo...
Read more

Data Mining and R/Rattle : First Experiment

-
Data mining is the activity of harnessing useful insights from vast amount of data based on a certain model. Instead of one particular model or way analysis, data mining is context-based and may potentially employ multiple models and can work on a variety of sources of data including text, audio, video, images etc. Data mining is an ongoing process and there are scenarios where the development of the data model stagnates are relatively few. This is because the mining insights act as feedback for a better model and hence, generally the process is much like agile software development. The role of a data miner in an organization begins with an understanding of the domain and data itself. This process is crucial to defining and refining the model that is the heart of the mining process. A model is constructed and critiqued by domain experts and data experts. This completes one cycle of data mining and the cycle continues in the same way every time using the insights generated from th...
Read more