This project was made for HIST3907B, Data Mining and Visualization with Dr. Shawn Graham at Carleton University. This website represents the culmination of my work in that course and how I have applied digital methodologies in order to do new historical research.
My project on residential schools was intended to examine public consciousness around residential schools through how people were searching for them on Google. While this was a simple approach, using it allowed me to uncover interesting patterns.
As I said above, the tool I used for this project was Google Trends. Google Trends is a feature provided by Google that allows a user to explore search trends over a period of time, beginning in 2004. While it appears that Google attempts to market it to average users with examples such as search patterns on Game of Thrones, it has become an interesting tool for historical analysis.
Along with discussions in class, I was inspired to use Google Trends to explore public historical consciousness by Shawn Anctil and Ian Milligan. Shawn Anctil is a PhD student in Public and Digital Histories at Carleton, and he has done lots of work with Google Trends - his Twitter can be found here. Additionally, Ian Milligan wrote about using Google Trends to explore Canadian public consciousness as related to the War of 1812 and commemoration in general - that piece may be found here. These two historians use Google Trends to explore history in unique and illuminating ways, and in my project I tried to evoke some of what they were doing.
Google Trends as a tool is hard to use due to it's normalization of data. A historian cannot access the raw search data that Google uses for their Trends visualizations, but instead normalizes that data as a value from 0 to 100. The highest relative searches in an area or at a certain time are given the value 100 and all others are assigned values according to their closeness to that. This makes using Google Trends problematic, as Google's algorithms are in control of delegating another level of artificial labels to the data, obscuring what can be discovered with it. For example, in my project I noted that it looked like the Northwest Territories had the highest number of searches about residential schools, when in reality that visualization was most likely from their small population and relatively high Indigenous population. However, this normalization makes the data easy for the average user to engage with and allows for easy manipulation and visualization. The graphs created through Google Trends would not have been as user-friendly or aesthetically pleasing if they were encumbered with huge values that varied greatly. In this sense, I am appreciative of the normalization process, though it would be useful as well to compare these normalized values to the raw search data.
In terms of the actual project, I felt that it all came together very solidly. Patterns emerged fairly easily when examining the data and the methodology of Google Trends allowed for patterns to stick out better due to normalization. I am very pleased with the conclusions I was able to draw from the data, such as the hypothesis that students are the ones most likely to search about residential schools, and communities with higher percentages of Indigenous population are more likely to be searching for information on residential schools.
I also feel that my visualization came out very nicely. While not as interactive as many of my fellow students' projects, I feel that using Google Trends scripts within a larger narrative exploring what those scripts showed was powerful. While there may be issues for some browsers in loading the scripts properly, I have added a note saying the project will work best in Chrome, which is a good compromise at this point I feel. I used Twitter Bootstrap and a template from Start Bootstrap to build both this document and my project, both of which are open source and free for anyone to use. While tinkering with the CSS and HTML was at times difficult, I experimented and found interesting solutions to problems.
Some of the conclusions I came to at the end of each section were definitely tenuous, and unfortunately not based in other documents for verification. I wanted to use only the Google Trends information to illuminate new patterns, but doing so forced me to make subjective guesses as to the meaning of the patterns I uncovered. Some patterns are fairly solid, such as my idea that it is mostly students who are searching for residential schools, and that searches are higher in areas with large Indigenous populations. However, my theories on semantic difference are very tenuous, and, while interesting, are not rooted in any outside sources.
Luckily, this is not a problem that takes away from the message of my project. In conceiving this idea, and stated in my introduction to the project, I wanted this project to act as a gateway for other historians to begin looking at public consciousness around Indigenous issues. Perhaps in the future I will come back and examine public consciousness around residential schools between central Canada and the prairies using educational materials in parallel to Google Trends. Or any historian is welcome to do so as well, as long as they cite my ideas.
One technical issue that I had with my visualization that has still not been resolved is the photo at the very top of both this page and the main one. The photo of the waves crashing on the rocks was included in the template I used for this site, and I had an idea of using a picture of a residential school in that introduction header to set the tone of the project. A black and white, gloomy photograph of an imposing building would set a negative tone for any user engaging with this project - as alluded to in my introduction to the main page, this is not a topic that should incorporate good feelings but instead should be a somber reflection. I cannot for the life of me figure out how to add the appropriate gradient to a photo I like in Photoshop so there is an even transition between the header and the black of the page. This is something that I may fix in the future, but for now is not the most consequential. It is a shame that the photo in the header has no real significance towards the project, but a necessary evil with my limited technical skills.
Update: after further tinkering with Photoshop, I have figured out how to do what I wanted with the gradient. Photo: Aboriginal children in class at the Roman Catholic Indian Residential School, Fort George, Quebec, 1939. Photography: Archives Deschalets.
Overall, I believe there was more that worked in this project than didn't. I hope to work out the kinks in cross-browser performance of the site, and make sure that the Google scripts load every time, but otherwise I am very happy with the outcome of this project.
This project veered very far from what I expected it to be in the beginning of this course. I originally wanted to incorporate analysis of both Reddit and Twitter posts to look at what was specifically being said about Stephen Harper's apology in 2008. However, problems with the size of the Reddit files and using twarc led me to beginning with an analysis of Google Trends. When I looked there, I found not only what I assumed I would find, that there would be a spike in activity around the date of the apology, but a wealth of other data and patterns that I have used this project to explore.
So instead of an analysis of one event using various tools, this project became an analysis of public consciousness over time using just one very specific tool. And, while I did not discover too much on Harper's apology, I instead got to analyze the much larger idea of residential schools in general, and how they were being discussed around Canada in the last ten years, not just in 2008.
My end result became very different than what I originally intended, but I think it is safe to say that this project achieved more. Looking at patterns in search over time, I have illuminated more patterns than I would have looking at just one event in various ways. While Reddit comments and tweets may have given some interesting information regarding the apology, I can assume that Google Trends would have given me very little when just framing the apology. Taking Google Trends into a larger context unlocked the tool's full potential for my project.
Thank you for taking the time to engage with my project! It means a great deal to me. Thanks to Dr. Graham for facilitating this work throughout the semester, and if you are a Carleton student that is at all interested in digital technologies, history, or data journalism, I fully urge you to check out HIST3907B: Data Mining and Visualization, in the coming year.
Ryan Pickering