DCI tools

Although I have not finalized the visual aspect of my digital essay yet, I have a good idea about what I want to do and have already started playing around with it. My hope is to compare various different articles and papers that specifically provide an opinion on AI in medicine. I have come across numerous ones that have either supported or been very against AI being used, so my hope is to show these different opinions through text analysis. Something like the voyant software allows me to take any article and run it to find the most common terms used throughout. From there, I can create a word cloud that will display these terms and can be compared to word clouds from other articles or papers. I hope by creating this visual there will be a clear understanding of the views that currently exist surrounding AI in the medical field. I hope to expand further on this in the actual written part of my digital essay, but this will provide viewers with the opportunity to briefly grasp an idea of where I’m going with the topic. I have tried searching for other softwares that do similar things but voyant seems to be the most user-friendly. I can literally paste a URL to the article or copy and paste the article directly in to a text box, and the tool does the rest. I have to play around with it a bit more but I am hoping to be able to specifically target certain terms so that the common but likely irrelevant terms, are left out of the word cloud. I specifically am looking at the trust or lack thereof among physicians in regard to AI so I am hoping to focus in on the terms that will be relevant to this.

Dear Future Students

To future students of DCI 101-

This class has been nothing but engaging and I am constantly learning new things that I would not have otherwise come across. Having already taken DCI 180 in the fall of my freshman year, I knew that this topic interested me, but it always amazes me how much more is to be discovered on the daily basis. DCI is one of those classes that is constantly being updated as new advancements are coming out, with the latest popular concept being Artificial Intelligence, something we have touched on a ton. I advise anyone taking this class to use every assignment to become more engaged in the world surrounding you. There is so much that exists that is waiting to be discovered and this class allows you to do just that. One thing I found to be particularly nice, is that there are so many directions that the course topic can be taken in.  For instance, our final project digital essay has allowed us to explore the part of DCI that most interests us, which I think is important with so many different majors in the class. This is one of those classes where you can really take in any one of your years at W&L. Currently, it is very senior-heavy but I think that is simply because we are almost, if not already, done with our majors and have the time to explore other interests. Overall, I highly recommend this class and have gained so much new insight from taking it, even beyond what I learned by freshman year.

Mapping the Coeducation Report

Above is the map I created using the coeducation report data we collected in class and the ArcGIS Online mapping tool. This specific map is pretty similar to the one I looked at for our class activity. It allows you to click on a specific points on the map and it will tell you the name of the person, where they live and whether they were for or against coeducation. This tool was free but required an account to use the mapping feature. It says the basemap is powered by esri which is a geographic information system company. Adding data directly to the tool was fairly simple. I was able to directly upload the csv file and the tool did the rest. I wasn’t super familiar with how to use the website to make the map more creative and interactive but I was able to create a map that had individual pins, representing the different alumni from the report. This map in general would be very useful for anyone wanting to create an interactive feature where different areas throughout the world or country correspond to certain people, events or landmarks. This was similar to the map I saw in the project I analyzed in class on Baltimore’s neighborhoods. Similar to this map I created, that map also enabled you to trace the different pins to learn more about the areas of Baltimore.

Spatial project review

This project known as “Baltimore Traces” is a teaching innovation that is a collaboration of multiple classes at the University of Maryland Baltimore. The media, which includes things like podcasts, digital maps, films etc. focuses on Baltimore residents and neighborhoods and is ever evolving as more aspects of the project are added. The project map they include on the website allows you to explore the different neighborhoods. There are certain tags on the map that allow you to identify different historical landmarks and events throughout Baltimore. For instance, I clicked on one part of the map, and it led me to information on the “Viva House” which is a soup kitchen and food pantry that was founded by a Baltimore couple 50 years ago. There is not only a written description but there is also a video linked to an interview of the couple. I don’t see explicit research questions, but it seems that the main focus of the project is to explore the different Baltimore neighborhoods, highlighting important things that have happened over the years in each. The map they created was made with Google My Maps, which is linked on the website should you want to create your own. I think the overall design of the website was well thought-through and easy to navigate. I can find exactly what I want to examine using the numerous sub-tabs that appear on the website’s main page, so I can imagine it is an easy tool to use for the general public.

My review was pretty similar to the review provided; in that I got the right idea about what the purpose of the project was. The review did however have more information about the aims, describing how the project specifically looked at deindustrialization, redevelopment and gentrification, which I didn’t pick up on. The review also went further to talk about the cross-institutional collaborations that made this project successful. The review also talks about how this project makes a significant contribution to Baltimore as it makes something that is very accessible for a broader audience, which I touched on in my review a bit.

Conference Recap

I had my conference the same day our digital essay proposal was due, so I had a pretty good idea of what I wanted to do. I explained how I had come across a lot about trust when researching about artificial intelligence in medicine, so we talked about making this my focus for the essay. This would entail looking at different sources to compare the opinions of multiple physicians regarding whether they trust the use of AI in their own practices. In addition to examining the opinions of those that are or would be using this machine learning software, I talked to you about how I also wanted to look at what would have to be done to ensure AI could be safely used with patients and receive the trust needed by physicians. We talked about the specific perspective I wanted to take and concluded that I would look at the overall topic of AI from the physician’s view rather than the patient’s view. This enables me to make a very broad topic more specific, and this interests me specifically since I hope to be working in the medical field in the near future. In regard to the visual aspect of the digital essay, I was unsure what to do but we talked about doing a text analysis and potentially using words associated with “trust” to compare different sources. The end product would be multiple word clouds that a viewer would be able to easily compare. Overall, you thought I had a good plan and didn’t think I needed to make any major changes to my proposal.

Digital Essay Proposal Draft

The topic of my project is artificial intelligence within the medical field, specifically looking at how it could be useful to physicians. Some research questions I have related to the topic include: how could AI change medicine for the better? What will AI have to entail to ensure it is a safe option for medicine, and not something that could go wrong very fast? What will need to happen to get physicians on board with trusting and utilizing AI in their everyday workdays? Who is responsible for AI malfunctions should medicine move forward with it? These are all things that I am hoping to look at with this project.

I have researched a decent amount about this topic in general and have come across some great sources, including ones that show when AI has been beneficial to medicine and ones where it has gone wrong. I think it is important to look at not only the benefits to using AI but also investigate its potential downsides. There is one article I came across that discusses the work of a physician at Mass Gen who is looking at how AI could be used as a clinical assessment to determine a patient’s cardiovascular risks based on just a chest X-ray. This is an example of a physician who is excited about the future of AI in medicine and is looking to incorporate it into his own practices. He also discusses the importance of the manufacturing process of these AI systems to ensure that they avoid error. Another article also discusses how AI could be useful, specifically looking at how it could be used to alert clinical staff of possible sepsis. It discusses the difficulties they have already had with that specific system and the alterations that would need to be made for long-term use. I have another source that looks more into the challenges of AI and the lack of trust from physicians, so this would enable me to look at what needs to be done before it can be trusted. There is an additional source that also looks into trust and challenges of AI. These are just a few of my sources thus far but I think this is the direction I’m heading in in terms of what aspect of AI and medicine I’m interested. I think it is an incredibly interesting topic and could be important to the future of medicine so am excited to learn more about what medical AI currently exists. I am not sure entirely sure what digital medium I am hoping to use yet but potentially mapping if I come across something that may work nicely with that. I am open to suggestions for that aspect of the digital essay.

 

Tentative schedule:

Week 9- continue research and finalize proposal

Week 10- determine how to organize digital essay and start working on it

Week 11/12- create analysis and written part of essay

Text Analysis Project Deep Dive

This text analysis project was created by Yale University Librarians Peter Leonard and Lindsay  King, both of whom come from very different backgrounds. Leonard is involved in Digital Humanities research, while King is an arts librarian with an interest in fashion. Bringing together both of their interests, they created this project that united two quite disparate topics, fashion and data mining. The corpus here is Vogue Magazine, in which through this website, individuals can explore many different questions that arise from it. For instance, one page on the website explores the different covers of Vogue over time. It allows one to visualize continuity and change without having to close-read every issue of Vogue front to back. Statistical methods have been used here to suggest themes and topics that have been consistent throughout time. There’s a page on the website where a program “reads” all the articles ever published by Vogue and produces a word cloud to show the most used words across all the magazine editions. There is another page that will separate all the magazine covers ever produced by Vogue into color hues, saturations and light variations.

So much information is condensed incredibly quickly by these computer programs that filter through all the data to produce an end-goal. I interpreted the research question throughout this whole website to be how data from Vogue can be quickly analyzed into any relevant theme. It is clear they wanted to form programs that would do just this, to prove that data can be quickly interpreted without the need to close-read every detail. In terms of the type of text analysis, there are multiple different ones being used. As I mentioned before, they used word count and topic modeling to produce results, but there are also numerous other types being used for the most holistic approach. Overall, I find the design of the website to be very appealing, providing a source that is incredibly easy to use and is intriguing to explore. It’s amazing how they really did essentially create “robots” to read and interpret every issue of Vogue magazine. It makes me wonder how many other things this could be used for if it works this well with just a magazine.

Coeducation Data Viz

Through this process, I have learnt that I am not the best at dealing with data. The condensation of the information into a spreadsheet was not bad, however producing visuals based on the data was a bit difficult. We came across the problem that not all the data was numerical as there were some columns like “support for coeducation” that consisted of just “Y”, “N”, or “A” referring to yes, no or ambivalent in reference to whether they supported coeducation. We were able to produce two simple visuals using a pi chart and a bar graph. We had difficulty using google sheets to show anything more complex so both visuals show the support of coeducation among the data we had, as seen below.

Pi chart showing the percent of alumni that supported, did not support, or were ambivalent to coeducation.
Bar Graph showing the percent of alumni that supported, did not support, or were ambivalent to coeducation.

After attempting to explore with google sheets, we imported our data into “RAW Graphs” which we found to be much more user friendly. With this website, we were able to create a visual that incorporated more than one variable. The x-axis represents the decade that the alumni graduated W&L, while the y-axis represents the average number of pages written in their letter. I spent a while trying to figure out how to create axis titles and a title but was unsuccessful in finding where on the website I would be able to do this, so that would be something I need to look into more in the future. Overall, despite not being able to easily find these features, we were more successful at incorporating more than one variable into a visual when we used the website rather than google sheets.

Bar graph displaying the average number of pages of the alumni’s letter for each decade the alumni graduated W&L.

I would not say I am a master at data but I do think if I play around with more data sets, I would become more comfortable with it. Also, while we did not think about data visceralization while we were creating these visuals since we were trying to keep it basic, I do think that there are ways that you could present the data to make it something more than just numbers. Coeducation is a topic that for many people evokes emotion so I think if you presented the data with certain colors or brought images into the visual itself, you may be able to make the data more meaningful. Just even through our pi chart, it became incredibly obvious how much the alumni were against coeducation. When you just read individual alumni letters you aren’t able to truly see how big the support was for keeping women out of W&L, but as soon as you combine these opinions into a visual you can clearly see that it was 66% of alumni that lacked support.

Website Genealogy

This is a website for PINC AI, which is an artificial intelligence software that uses premier technology to deliver smarter and faster healthcare. The site allows you to explore the different features of the software and enables you to see what types of work it could do for the future of healthcare. You are able to click on different icons throughout the website that tell you more about the AI tool and enable you to see all of its different features that could potentially make life much easier for physicians. The site is pretty straight forward and easy to navigate, but at the same time it is very visually appealing and encourages the viewer to want to keep exploring. The overall vibe of the site is very futuristic and it makes the concept very exciting as a viewer, looking on to this new piece of technology that could be the future of medicine.  

.

I used the way back machine and I was able to follow the website as it was being built. It was first published in October of 2021, however it was still in its earliest stages then. The font was incredibly large and there were no interactive features that now exist on the site. By March of 2022, the site was much more put together and started to look more like it does now. Interactive features started to appear and font color and size made it much more appealing to look at. While the actual content did not change an incredible amount from 2022 to now, the appearance of the site as a whole changed tremendously. The site only grew to be more visually-appealing and further structured in a way that makes it very easy to navigate. There are clear headers that take you to different pages which share features of the AI tool. Each page has multiple different tabs that can be clicked on for further information and analysis of how it works. Overall, I think this website is incredibly successful at sharing information in an informative but stimulating way. It is hard as a viewer of the site to not want to continue exploring since there is so many visuals and text to interact with. 

Data Viz

This data visualization conveys how artificial intelligence (AI) and machine learning (ML) impacted healthcare in 2022 by displaying its effectiveness across multiple categories. I think the presentation of this data is very clear through this visual and shows a good representation of AI’s impact on the medical field. The visualization was produced by Insider Intelligence which is a market research company, and the data was supplied by a software company known as Innovaccer. I tried looking for the original data set used to produce this visual but was unsuccessful in my findings. I think the audience here would be the general public who may be curious about how effective AI is in the realm of medicine. An individual to come across this visualization would easily see the effectiveness on AI through the opinion of healthcare executives.

It is possible that this data is biased as it is solely looking at the opinions of those in the healthcare field so depending on whether they feel AI has a place in medicine they will likely answer accordingly. Those who are excited about the role of AI and the benefits it could have would likely say that it was effective regardless of whether it actually was. However, those health care workers who are strongly against the idea of AI in medicine are going to be more reluctant to say it had been effective in their practices or hospitals across those six categories. I think it would be interesting to compare these results to the job title of those providing the data. By drawing this comparison, it may be possible to see a correlation between what kind of people are or are not supporting AI. Overall, I do think this visualization is effective as it is very reader-friendly and can easily be interpreted. Immediately as a reader, I can see that the data generally did not populate in either extremes of the spectrum but rather was contained to the middle, which was reflective of “often effective” and “sometimes reflective”.

css.php