The Monthly Briefing

We're not statistically minded animals



As countries came to a standstill while COVID-19 was spreading around the world, the pandemic has brought to light some things that had previously gone unnoticed. One of these is the importance of clearly communicating to people the results of scientific research or the implementation of certain social norms that need to be understood by everyone in order to be effective

Within our own modest world, we are also asking ourselves how we can best communicate the output of our algorithms to our customers, so that they can get the most value out of them. This is not an easy task: we humans are not used to working with imperfect results. Especially when it comes to evaluating probability-based scenarios.

This is what is known as the wet bias phenomenon: it has been observed that meteorological services in certain US (United States) media sources were used to deliberately inflate the probability of rain to be much higher than had actually been calculated. In his well-known book “The Signal and the Noise”, the statistician and data disseminator Nate Silver delves into this phenomenon and goes so far as to attribute it to the fact that meteorologists believe that the population, whenever it sees a probability of rain that is too small -say 5%-, will interpret it directly as “it’s not going to rain” -and consequently will be disappointed 5% of the time-. In other words, we humans tend to simplify information for decision-making.

For this reason, we must be very careful not to present the results of a forecast or a prediction as absolute, for example. In this particular case, it is more effective to analyze how confident the algorithm is in each forecast, and perhaps discard the cases where we do not have high confidence. This article by our colleague José A. Rodríguez Serrano goes into more detail about all these considerations


Another crisis that needs to be tracked

Via Getty Images.

On the way to changing our relationship model with the planet towards one that is more sustainable, governments, public and private institutions and society as a whole must be able to manage the emissions of polluting gases warming our atmosphere. And as David Roberts says in this article in Vox, if we want to manage something properly, we have to measure it beforehand.

To tackle climate change, one of the main challenges to be solved is to correctly track where greenhouse gas emissions caused by human activity occur. So far, this has proved to be a complex task, which in turn has made international climate negotiations extremely difficult.

However, the solution to this problem does not seem so far away. The Climate TRACE (Tracking Real-Time Atmospheric Carbon Emissions) Coalition “is building a tool that will use artificial intelligence, satellite image processing, machine learning, and other remote sensing technologies to monitor worldwide greenhouse gas (GHG) emissions in real-time”. This tool will be verified independently and will be available to the public for free.

According to Gavin McCormick, executive director of coalition member WattTime, “the Earth is like a medical patient suffering from a condition called climate change. Trying to fix it with only years-late, self-reported emissions data is like asking a doctor to fix a serious disease with no more information than a list of symptoms the patient had years ago”.

One of the most interesting future applications of this tool is to provide crucial visibility to more-easily and accurately meet emissions-reduction goals, direct sustainable investments (and divestments), and assess risk. You can find more information in this articles in Medium (1) (2).


Further reading

+Using Convolutional Neural Networks for image processing and classification: a case study on Sudan's Darfur region (Citizen Evidence Lab)
A very illustrative example on how to use the potential of artificial intelligence for large-scale analysis of satellite data to detect the destruction of human settlements. To do that, Amnesty International enrolled digital volunteers to scan through satellite imagery to identify remote villages in Darfur and determine if these had been damaged or destroyed between 2014 and 2016. After that, the aim was to learn the task of mapping an image to the correct label. In other words, given an unseen tile, we wanted to accurately predict whether a human annotator would mark that tile as a destroyed village.

+ What is a real data-driven decision and the danger of falling into a confirmation bias. (Hackernoon | Medium)
The key: setting your decision criteria in advance.

+Why the path to AI-first banking is a must (McKinsey)
This article by Suparna Biswas, Brant Carson, Violet Chung, Shwaitang Singh, and Renny Thomas outlines the major trends being observed in the banking industry that are charting the way towards central adoption of AI technologies.

The increased expectations of customers regarding the digital maturity of companies -enhanced by the COVID-19 crisis-, the benefits of adopting AI in parts of the business such as process automation, risk management, fraud detection or customer relationship, and the strong competition in the digital ecosystem, are new elements that banks are facing.

The bank of the future will be AI-first, it will offer propositions and experiences that are intelligent, personalized and truly omnichannel. However, for this to happen, traditional banks will have to overcome certain obstacles. Some of the most important are the lack of a clear strategy for AI, a weak core technology and data backbone, an outmoded operating model or the several weaknesses inherent to legacy systems before they can deploy AI technologies at scale.

The main features of the AI-first banking. Via McKinsey

+ Facial recognition ban awaiting regulation (c|net)
This is what a group of lawmakers has proposed until US Congress passes a bill lifting the ban. The moratorium would affect federal agencies such as the FBI, as well as local and state police departments nationwide. The reasons: this technology could amplify discrimination against people of color, and also deter people from exercising their free speech rights.

+ Is facebook amplifying the spread of the messages raised by the right wing? (The Verge)
The verge publishes many conversations that have taken place between Mark Zuckerberg and his employees over the past few months. They show the gap between employees and the user base of Facebook. Zuckerberg seems to trying to hold the center while criticism comes in about how the social network may be polarizing public opinion.

+ The GPT-3 revolution explained in this video (DotCSV) (in spanish)
GPT-3 is very powerful, but we have to analyse correctly what it can offer and whether it meets the high expectations. In this interview in HAI (Human-Centered Artificial Intelligence), Oren Etzioni, Allen Institute for Artificial Intelligence CEO and professor of computer science, explain that although "it is very remarkable what we can do with it, particularly at a superficial level, like generate a fluent text and so on, we have to be very careful to distinguish these impressive behaviors from a genuine intelligence". You can watch this interview on youtube.

+ Time to dataviz (Area Estratégica)
This is one of the oldest organizational structure documents in history: the New York & Erie Railroad Diagram. A great example of information design.

Plan of organization of New York and Erie Railroad. Source: McKinsey Quarterly

+ The magic of TikTok (TikTok)
This article explains how works one of the most valued aspects of TikTok's app: the recommendation system behind the feed.

+ Another interesting analysis on The Social Dilemma. (Slate)
Pranav Malhotra highlights in the article the utopian vision of the people within the big tech such as Facebook, Google or Twitter, but also puts the focus on the tendency to underplay how maximizing profits is a major motivating factor for these same companies. Not to mention that, although the documentary seeks to find the real issues on the tech industry, no mention is made of problems such as inequality or lack of diversity in this field, which also influences the design of these platforms. According to Malhotra, the documentary shows technology as the sole cause of complex phenomena that we should not analyse without considering all the factors that may be involved.

Quote

of the

month

"Chess engines were initially built to play against humans with the goal of defeating them. Now we see a system like AlphaZero used for creative exploration in tandem with humans rather than opposed to them."

Nenad Tomašev, Senior Research Engineer at Google DeepMind


AlphaZero is a more flexible and powerful successor to AlphaGo, the computer program developed by Google DeepMind that made history in 2016 by beating the world champion in the game of Go, Lee Sedol. Since last March, the complete documentary that tells the story of this tournament is available on youtube (highly recommended).

Now, Vladimir Kramnik, a former world chess champion, teams up with the makers of AlphaZero to test variants on the age-old game that can jolt players into creative patterns. More in this article.

BBVA DATA GALAXY 🌌


Data is vs. data are...

We would like to finish this issue with this amazing little comic. Sometimes the silliest question generates the deepest debate!

Thanks to Luis Sainz for the reference ❤️. Via PhD Comics.

See you next month!

< Read previous issue

For any question or suggestion, you can also write to hello@bbvadata.com

You can enjoy much more content related to data science, innovation, new solutions of financial analytics and how we work in our website: bbvadata.com

Let's talk about it. Join the conversation on Linkedin.

© 2020 All rights reserved. BBVA Data & Analytics. Avenida de Manoteras, 44, 28050. Madrid.