Towards peace through exposition of differing perspectives

According to Alexa traffic rank, Google is the number one most visited website in the world as of August 2015. But as much as Google pledges to be a source of open information, let's think for a moment: when was the last time you were on Google without a topic in mind already? Never. Because search engines simply don't function that way.

You always begin your search with a pre-primed mindset that is reflective of your culture and background, as well as your current environment. As a result, the natural question that continues without solution today is: how can you discover if you already know what you're looking for?

Even after your search result populates the screen, the stream of narrow-mindedness unconsciously continues. Google's algorithms display search results in a deterministic way. With the average user never venturing past the second page of results, the world of information online is really only limited to what Google presents to you within the first few pages--the links that most match your search prompt. Hence, your results will always be things you're already looking for.

People only see what they want to see by searching for what they want to find. Algorithms today are becoming too specific by targeting you through your demographic and displaying exactly what you want to see. As a result, you are only finding what you want to find, not discovering new information and new perspectives into an issue.
Specifically, the lack of discoverability of the current search engine structure can be attributed to the following reasons:
  • Subjectivity of reported information dependent on region
  • Censorship of information dependent on region

Take a look at the current search flow:
  • For each topic you’re interested in:
    • For each newspaper you already know about:
      • Type in the search results on google.
      • Select some link to read (which link? Most likely random, or from within the first 3 results)
      • Try to understand from which viewpoint the article is written in

The search results presented will be based on your location and language, and what other information Google thinks you want to see. Your selection of which article to read is relatively arbitrary (demonstrated by user testing). This is an inherently biased result.

Let's illustrate this problem with a concrete topic: the most recent news of the attacks on Paris. Upon seeing statuses on Facebook, radio, or the news, a person will go to Google, and type "Paris Attacks." The top search results are from English outlets such as BBC news and CNN. But, if instead, s/he googles "ataques París," the top search results are from Hispanic outlets, such as El País or El Tiempo. The search results that Google presents are language and region dependent. But the perspective that El País presents may be different from the perspective of BBC, or that printed in Kitabat, an Iraqi newspaper.

As another example of subjectivity of reporting dependent upon language and location, take a look at the Japanese earthquake of 2011 of Tohoku. In reporting these events, foreign newspapers often dramatized the situation, using exaggerated words to describe what happened (the Daily Mail used the title"Japan's Apocalypse Now" on its March 16th 2011 article). International media also focused on the efforts of the Japanese to rebuilt their country and kept showing their solidarity: the Independent's Sunday cover had the title "Don't give up Japan! Don't give up Tohoku!". On the other hand, Japanese media was much calmer and mostly diffusing the information given by official sources from the Japanese government and TEPCO. While foreign media such as the New York Times or ABC News were praising the courage and selflessness of the Fukushima 50,( , Japanese newspapers did not give major coverage nor glorify these workers. Asahi Shinbun, one of the major newspapers, also covered the rising problem of donation scams and theft that was going on after the earthquake.

As shown above, even the most well-intentioned, neutral person is receiving filtered information based on language and location in which the search is conducted. How can we truly understand an issue if we only see it from the one perspective we are already familiar with?

The path to peace is understanding other people’s perspectives from around the globe. While the goal of world peace is lofty indeed, this project will aim to foster understanding by gathering and presenting existing data in a different way to enhance discoverability.

external image mpj0430642000011-e1316520542392.jpg

Hypothesis (Solution)
The solution we propose is an extension or a reworking of Google's existing Trends product. Currently, Google Trends presents the most popular searches given a category and a location. However, this is not useful for comparisons of different perspectives regarding a specific topic.
Instead, we propose an online platform that gathers information from different regions about a specific issue/topic by:
  • Showing which countries or regions talk/do not talk about the issue
  • Showing the popularity of this issue in this country (as measured by the number of searches)
  • Showing the different perspectives that local media is taking, as a variable of region

Proposed userflow:
  1. Home page, with a search bar. User types a search term in the bar
  2. A world map displays, each continent color-coded, with what this continent is saying about the topic. For each differing opinion, present the opinion in a succinct 200 character summary blurb.
  3. Click any blurb to see newspaper articles from that region regarding the topic. Each blurb will be a link to their respective articles.
  4. Click on any region to zoom in. Then region specificity increases
Screen Shot 2015-12-04 at 9.12.46 PM.png

This new method of searching makes comparisons of global perspectives intuitive and straightforward. Before, you had to compare search results from different regions by explicitly amassing the information yourself ("Paris attacks CNN" and "ataques París El País"). Now, the comparison is visual and easy.

Is geography really the most characteristically relevant in identifying differing viewpoints?
Possibly not. Therefore, in addition to the default geographical search, we propose creation of additional filters by religion, ethnicity, GDP, size of country in terms of population and landmass, political tendencies, and other defining characteristics of a region from which the newspaper is based. A similar display process will ensue. The visualization would be adjusted for each filter. For example, suppose search is performed by religion. Then, instead of a map of the world, there will be symbols representing each major religion, with a summary blurb as described above with newspapers originating from regions that are predominately that religion. A click into each major religion will split the religion in to sects.

  • Use google trends to amass trending topics. Pick headlines that are trending within all countries
  • For each trend, get links to all newspaper articles written about the trend
  • For each newspaper, map it
  • For each article, perform a sentiment / keyword analysis
  • Visualization of general sentiment of each continent / country / region regarding that article
  • Archive search results to see the sentiment changes across time, visualized in a sliding timeline
  • Language: will support multiple languages by translation. Some newspapers have official translations of their articles. Some do not. For those that do not, use a translation engine, with a disclaimer that translations might be inaccurate

external image Sentiment.jpg

As it is a website, we can employ the typical measures of website success with Google Analytics and surveys, such as:
  • How many active users / unique page views per day
  • Survey to users for feelings before and after on an issue. Did this website change your view point? Do you now understand the different viewpoints from which you can see this problem? Does this make your research life easier?

external image shot_1398345779.png

Evaluation and Challenges
At a larger scale, the goal of this project is to try to foster understanding between opposite corners of the world. As such, there is no quantitative measure for the success of this solution other than finally seeing world peace. At a smaller scale, this product aims to be of tremendous use to scholars to understand the complexities in our increasingly interconnected world. As an evaluation of this project overall, if this project provides utility to some subset of people, then it is considered a success.

As with any experiment, there are challenges. In particular, the data processing and sentiment analysis to generate succinct 200 character summary blurbs is one of the largest perceived challenges, especially with the added factor of language differences, and even within the same language, regional differences in phrase usage. Disregarding the complexities of linguistic analysis, how will the visualization tool handle discordant opinions within the same geographical region? Currently, the idea is to provide a different summary for each different opinion. But what if, within a region, there are many different opinions such that it becomes psychologically distasteful and/or impossible to go through them all? This is especially applicable at the first search result, with the map of simply continents. One cannot expect the whole of Europe to only contain 5 different opinions about ISIS, for example. And further, what is considered an opinion? In a world of so much information, how do you generalize meaning from some assembly of words, and then report the subjective opinions of each article accurately in a relatively quick run time?

What about the other filters? The current product stipulates that each newspaper be categorized according to the defining characteristics of the majority of the population in the region in which it is printed. For example, a newspaper would be categorized as Catholic if it were printed in Mexico, where Roman Catholicism dominates. This obviously creates some inaccurate results. We will need more reliable classifiers for differing perspectives on a topic.

Propose solutions
Here, we present solutions to some of those problems listed.

We can lessen the burden of different languages by only sourcing newspapers that have English translations of their articles (or the 1 standardized language that this tool operates in). Newspapers have an incentive to provide a translation so that their article can be listed within this tool, driving more traffic to their website.

Regarding the issue of too many discordant opinions, we can employ UI tricks (such as hiding some search results within a "next" button) to make the result more visually accessible.

Q: What new data would you use? How would you obtain it? How would you use it?
A: There is no new data being collected, but rather presenting existing data already in Google's databases in a more usable way. You will use search data from Google’s daily searches in order to view what other people are thinking around the world.

Q: How will this new data improve the end user experience? Clarify how you can really improve the product.
A: You’re improving the search experience. You’ve always searched by topic to find specificity about this topic. But current search results are heavily biased based on region and language. For example, regarding ISIS, is western media reporting the same thing as other media centers around the world? Unless you specifically look up this topic in a (for example) Chinese newspaper, you’re only going to get results from Western media outlets (CNN, NPR, Fox, BBC, MSNBC). Current results don't have a global scale, which we need in an increasingly interconnected world. You’re discovering instead of finding what you are predisposed to think based on your cultural preconditioning.

Q: Why should I care? Why should Google implement this proposal?
Google's mission is to organize the world's information and make it universally accessible and useful. Google does very well with giving you more information about a topic you already know something about. But currently, there is an unmet need for what we have termed "upwards searching" --or discoverability of information you know nothing about. This project aims to make information more discoverable.


Mona Ito, Yuan Yuan