UW iSchool Course INFO 474: focused on designing and building visualizations to better understand and communicate about pressing issues.
md`
### Level 1: Characterize the tasks and data in the vocabulary of the problem domain.
The first level essentially tells the designer of the visual to figure out what area of knowledge their data and subsequent visualization falls under. The designer should know the problem they are attempting to address inside and out while showing users the most compelling and effective data that solves or displays the problem. Users should not walk away confused because the designer themselves was confused as to what domain the problem falls under.
### Level 2: Abstract into operations and data types.
From my understanding this is the point where designers think about and build the objects and relationships that are the foundation of the visual. This includes transforming raw data into data types. In the paper referenced there is talk of developing a taxonomy, and therein lies the heart of what the second level is; a categorization of data into data types, objects, and operations that are used to construct the code base.
### Level 3: Design visual encoding and interaction techniques.
This is where the magic happens. After the hours of prep going into cleaning and transforming data it is neccessary to think about how the visuals will appear and what their level of interactability should be. Will it be a bar chart race or a static word cloud? These considerations are made here and planned out to the smallest detail.
### Level 4: Create algorithms to execute these techniques efficiently.
This is the nitty gritty of what makes your visual run and display visuals in a timely and clean manner. With a slow running algorithm an animated visual of a map situated in the US that displays where people are infected with COVID-19 might not even show the fill of colored dots of infected areas. Whether it be the complexity of the algorithm or the running processes of a desktop computer a designer needs to take all the barriers-to-entry into account.
`
md`
Level 1 Output: Really specific questions that are derived from the domain of the problem but attempt to address smaller components of the context the data falls under. This helps for deciding what types of data the raw inputs of the first level should be transformed into for the second level.
Level 2 Output: A set of operations and data types that are used to start building the visual. After securing the raw data and transforming it into a series of objects and data types the sets of data are moved into the layers of the visual itself.
Level 3 Output: To put it simply, the output is the visual itself and whatever interactivity is associated with the visual.
Level 4 Output: The power to display and run the visual(s).
`
md`
Level 1 Threat: Designers tend to make assumptions about what they see in the data and that bias cascades down to implementing the algorithms that end up displaying the data. Opinion and bias are threats to the first level.
Level 2 Threat: A designer will typically skip the first level of domain characterization and head straight here. This is a threat to this level because without proper consideration of all aspects of the domain a designer will often select the most interesting object to focus on and forego other important data types and operations that could be found in the raw data.
Level 3 Threat: There is a myriad of ways to develop the visual encoding and level of interactivity. The biggest threat at this level is when a designer picks an ineffective technique to use without considering the limiting aspects of an inferior technique.
Level 4 Threat: In my own words, I would say the biggest threat at this level is that designers get lazy and create alogirthms that run slow and don't capture the entire set of operations needed to effectively display their objects and data types built in previous levels.
`
html`<img src='https://www.mastersindatascience.org/wp-content/uploads/2014/03/U.S.-Gun-Deaths.png' />`
md`
<p class="subheader">Domain Problem and Data Characterization</p>
1. What **domain** is the visualization coming from?
The domain is the topic of gun deaths in the United States of America.
2. What domain specific **vocabulary** is used in the work?
This visual doesn't utilize too much vocabulary in the visual itself but two instances of diction stand out in particular. The "502,025 Stolen Years" card in the upper right corner illustrate the combined value of years people held that have lost their life to gun violence. The other vocabulary use I noticed was the categorization of "Young Adult" to include everyone aged 0-30.
3. Who is the pertinent **target audience** within this domain?
On a large scale, everyone that resides in the United States. However on a smaller scale, I would say the target audience is young adults or policy makers in America.
4. What **specific questions** does the target audience hope to answer?
How many people have lost their life to gun violence in 2013? What percentage of that population are young adults? What is their sex, region or time they were killed?
`
md`
<p class="subheader">Operation and Data Type Abstraction</p>
1. What generic _operations_ are required to answer the question(s) identified above? Indicate whether each operation is a _high-level_ or _low-level_ task, as described by Amar and Stasko (in the Munzner paper).
The operations would have to include high-level tasks such as how to categorize the groupings of age and sex as well as low-level tasks such as calculating percentage from a total numeric value.
2. What are the data types present in the data (e.g., categorical, continous, etc.)?
This visual includes both categorical and numeric data types. The categorization can be seen in the groupings of age and sex. The numeric data is most easily seen in the compounded data of years stolen from the populaiton (502,025) and the lives lost (11,419).
`
md`
<p class="subheader">Visual Encoding and Interaction Design</p>
1. Which visual encodings and interactions were selected to support the tasks identified in the previous step?
I'm actually not sure how to answer this question as I really do not know what type of visual is displayed above. However, this is also the reason why I selected this visual in the first place. The main visual appears to be a colored bar that breaks out into individual threads colored at specfic points along their lengths. This abstraction makes it easier for user to visualize the entirity of the amount of people who lost their life due to gun violence. I also love how effective the large numeric numbers are in the two upper corners as they both answer the two most prevelant questions in a user's mind when they navigate to the visual. They are the two questions that first come to mind when a user considers the domain of the visual.
2. We'll discuss this more in next week's notebook, but based on your current knowledge, do you believe these encodings and interactions allow users to sufficiently accomplish their tasks?
I honestly do not love the visual. The image it leaves in the user's mind is striking but doesn't convey the complete message of what I personally want to know. It is also difficult to display the different categorizations in the tab at the bottom without seeing each categorization displayed on top of each other.
`
html`<img src='https://www.kaushik.net/avinash/wp-content/uploads/2017/f54bdac20c6b_949E/mapping_slavery.png' />`
md`
<p class="subheader">Domain Problem and Data Characterization</p>
1. What **domain** is the visualization coming from?
The visual illustrates how the demographic of black people in the U.S. are spread across the varying regions with the added element of displaying the federal and state incarceration per 100k people in that specific location.
2. What domain specific **vocabulary** is used in the work?
State prisoners, slavery, black population and state.
3. Who is the pertinent **target audience** within this domain?
On a large scale, everyone that resides in the United States. However on a smaller scale, I would say the target audience is black people in America.
4. What **specific questions** does the target audience hope to answer?
What is the incarceration rate of black people in America? Is that rate substantially higher or lower than other demographics in the country? Where is the target demographic primarily located? Does region have anything to do with where a demographic resides?
`
md`
<p class="subheader">Operation and Data Type Abstraction</p>
1. What generic _operations_ are required to answer the question(s) identified above? Indicate whether each operation is a _high-level_ or _low-level_ task, as described by Amar and Stasko (in the Munzner paper).
High-level tasks would be categorizing numeric percentage values with their color orientation, alternating size with rate of people held in prison, and determining total size of different demograhpic populations compared to the total. Low-level tasks would be color type, determing the visual display of individual point markers.
2. What are the data types present in the data (e.g., categorical, continous, etc.)?
This visual includes both categorical and numeric data types.
`
md`
<p class="subheader">Visual Encoding and Interaction Design</p>
1. Which visual encodings and interactions were selected to support the tasks identified in the previous step?
The most striking choice made is the idea to overlay the data on top of a map of the United States. This gives users a sense of an orientation and how prevelent some of the numeric data is in certain areas of the map. It explains more with the image than it ever could if it was just on a spreadsheet. I also liked the use of color and size to illustrate different data types while still pertaining to the exact location of an area.
2. We'll discuss this more in next week's notebook, but based on your current knowledge, do you believe these encodings and interactions allow users to sufficiently accomplish their tasks?
This visual accomplishes a lot more than the previous visual in my opinion. The use of the map and intutive choice of utilizing both size and color for the individual point markers is ingenious.
`
Purpose-built for displays of data
Observable is your go-to platform for exploring data and creating expressive data visualizations. Use reactive JavaScript notebooks for prototyping and a collaborative canvas for visual data exploration and dashboard creation.
