Over these past few weeks, I have attempted to quantized some aspect of my daily activity to explore the relationship between data and their ‘insights’ within the context of learning.

I started the block thinking it would be an easy task – to record data and visualize it; but, I quickly realized that it would not be. Each week I faced a unique challenge of what and how to record data for the particular activity. In deciding the what, metadata was often omitted or reduced/regrouped as the week progressed, especially while drafting visualizations. Here we can begin to understand data as being partial and selective according to the context of its collection and end use (Michael, 2016; Williamson, 2017). As Kichin states, as cited by Williamson (2017), the term ‘capta’ should be used rather than ‘data’ as data are “inherently partial, selective and representative, and the distinguishing criteria used in their capture has consequences”.

The visualization of the data has an impact of what conclusions can be drawn. For example, the design of my first visualization, a week of walking, a chat log was emulated and messages were grouped temporally. Consequently, some of metadata that I would have like to have included was omitted to keep the visualization easy to read. Rather than grouping messages temporally, I could have grouped them by category (e.g. code, simulations, lab manuals, etc) or by sender. Breaking away from the chat log style, a series of pie charts or scatter plots could have been developed where there would have been more of a numerical focus. Or, all of these visualizations could have been developed to provide a range of perspectives of the same data set in the hopes to reveal a range of insights that might not be obtainable without the collection and analysis of the data, which is an aim of learning analytics (Knox, 2020).

In regards to the numerial focus, which seems to be the trend in the collection of student data, Bulger (2016) warns that students urged towards a quantified outcome will focus more on reaching that value rather than on the process of learning itself. While I agree with Bulger, I find this to not be unique to personalized learning analytics, but rather to most educational practices where grades are assigned, which is a form of reducing learning to a data point.

Target values are often predetermined and students may vary in their approach to a specific task subsequently leading to meeting, not meeting, or surpassing these values and lead to false positive or negative outcomes (Bulger, 2016). Knox (2020) highlights a new form of hypernudge platforms that ‘nudge’ students into these predetermined values. In these systems of realiging students to predetermined values or trajectories, student agency is attacked (Bulger, 2016; Tsai, 2020).

Throughout the data collect process I was the subject and data recorder and had a deep understand to the data that was collected, but unfortunately this is not the norm. As Tsai (2020) notes, the lack of full transparency of data and algorithms can lead to a distrust of the analytics and can further remove student agency preventing students from directly challenging the precision of the analytics. How data is collected, used, manipulated, and shared needs to be transparent and open to interrogation from educators and students – the ‘black box’ needs to be opened (ibid).

Bulger, M., 2016. Personalized learning: The conversations we’re not having. Data and Society, 22(1), pp.1-29.

Knox, J., Williamson, B. and Bayne, S., 2020. Machine behaviourism: Future visions of ‘learnification’ and ‘datafication’ across humans and digital technologies. Learning, Media and Technology, 45(1), pp.31-45.

Michael, M. and Lupton, D., 2016. Toward a manifesto for the ‘public understanding of big data’. Public Understanding of Science, 25(1), pp.104-116.

Tsai, Y.S., Perrotta, C. and Gašević, D., 2020. Empowering learners with personalised learning approaches? Agency, equity and transparency in the context of learning analytics. Assessment & Evaluation in Higher Education, 45(4), pp.554-567.

Williamson, B., 2017. Big data in education: The digital future of learning, policy and practice. Sage.

This week the US had the second impeachment trial for the former president Trump. I watched the entire trial through the CSPAN, the network that primarily televises the proceeding of the US federal government, live stream on YouTube. Additionally, I had the Twitter live stream opened. As the trial proceeded, I often had questions or wanted to seek clarification/more information from points raised during the trial by the lawyers/impeachment managers or commentary through Twitter. I thought it would be interesting to map my path throughout the trial.

Methodology

Rather than trying to document by hand all of my web activity, I relied on my browser history to record them. At the end of the day, I printed my browser history and I tried to label whether activity was done to answer a question, as reaction to Twitter commentary, or content algorithmically provided. This data resulted in the visualization below.

Visualization Design and Discussion

Each day of the trail, excluding Saturday, is represented by a purple circle which also represents the livestream of the trial and the Twitter live feed. Branching from the day are my web activities that result from watching the trial. They have been categorized as: YouTube videos, Tweets/Twitter accounts, printed news media (e.g. articles from NYT, CNN, and FiveThrityEight), video news media (e.g. clips from CNN and Fox), Google searches, and instant messages (e.g. Discord). In total, it forms a web for that day.

Some branches also include a square, a triangle, or dashed boxes; these represent notes for that branch of activity: question/personal inquiry (square), notification/algorithmically provided (triangle), or tangential queries not related to the trial (dashed boxes).

Looking over the week, I thought it was interesting that I did not perform any google searches on Thursday. However, after thinking about the timeline of the trial it makes some sense: Tuesday was focused on determining whether or not the US Senate had jurisdiction to hold the trial, Wednesday was the opening arguments for the impeachment managers, and Friday the former presidents council presented their defense. In each of these days, legal terms and precedent were introduced as well as the introduction of members of the defense and prosecution. On Thursday, the impeachment managers focused on video footage and summary of their case. Lost in this visualization is the context of the search and where I found the answer to my question. I felt that it was important to include the tangential queries as they represent learning that might not be considered as ‘focused’ or ‘on task’ by some.

Most of my activity seems to be from Twitter, again this makes some sense as I had the Twitter livestream commentary open the entire day. One side effect (or benefit) of the livestream is the automatic refresh of tweets and I did not count this refresh as activity. Twitter activity was defined as opening the tweet to view the comments or view the account that posted the tweet. I did not consider liking or following accounts as activity.

This week, I focused on my walking habits. During my walks I try to take a break from technology and reflect on my day.

I generally work from 09:00 – 18:30 and try to get two 1-1.5 mile walks each day: one in the morning (~11:00) and the other in the afternoon (~15:00). In addition to the data that I collected, I used Google Fit to track my distance, steps, and “heart points”.

Methodology

Google Fit: This data was recorded automatically from my phone; I do not use a wearable device. The data collected from the app is for all walking activity for the day – not just reserved walking time.

Personal: During my walks I recorded what my thoughts were primarily about – fleeting thoughts were not recorded. After each walk, I noted if I had been listening to music while I was walking.

Both of these data are displayed below:

Visualization Design

Google Fit displays data using concentric circles where the outer circle, green, represents “heart points” and the inner circle, blue, represents total steps – both of these are relative to set daily goas. I decided to emulate the circle style by breaking down my thoughts relative to the total walking time. I estimated the percentage of walking time by landmarks along my walking path. Other events, such as a phone notification, changed what I was thinking about these are marked by a single or double dash. The circles should be read counterclockwise.

Discussion

I have used Google Fit for a while, but I haven’t looked at the data too closely; usually I just look at my daily step count. It was interesting to look over the daily breakdown and see how the app classifies some of my activity and how the app assigns “heart points”. Some walking activities with similar distance and “movement time” gave me vastly different heart points. Some entries of the exact same walk resulted in different distances yet similar steps (differing a few 10s of steps). The walking paths displayed by the app are also a bit wonky and that prompted a comparison to my tracking data from Google Maps. There, I saw some of my walking entries in Google Fit were marked as “Missing Activity”, and Maps displayed a more accurate walking path. Maps was also better at filtering out miscellaneous walking entries since Maps filters by location rather than moments of increased activity (e.g. bringing in the groceries).

Also looking over the data from Google Fit also gave me a chance to update my daily goals which were set before working remotely where I rarely sat at my desk leading to higher goals. This is a reminder that goals need to be continually re-evaluated and updated with changing environments.

During my walks I try to take a break from technology, and I would say that I have mostly accomplished that: 6/9 walks I did not listen to music and I did not respond to notifications half of the time. I also try to use the time to reflect on what I have been working on before or what I should do after the walk. For example, I usually take my afternoon walk around the time I finish lunch where I will catch up on reading or review the blog, and for most of the afternoon walks I spend some time thinking/reflecting on school work. Finally, there also seems to be a trend where a notification, especially if I respond, disrupts what I was thinking about. This makes sense since I would have had to stop what I was thinking about to respond which leads to new a train of thought.

This week I decided to track the number of work related discord messages that I received. For some background, I work within a group of four supporting the physics undergraduate labs at the University of California, Riverside. Currently, I am focused on developing and converting simulations for our remote labs. We use a discord server for voice and text chat between the tech staff (the team I am apart of), select faculty members, tracking our git updates, and a text based role-playing game; not to mention the various direct message chats. For many reasons, this week was relatively relaxed with only 435 messages!

To track the messages I recorded: (1) who sent the message (bot, tech staff member, or myself), (2) when the message was sent (morning or afternoon), (3) was the message sent as a group message or a direct message, (4) and how I viewed or sent the message (desktop vs phone). There were no faculty messages sent in the discord chat this week. Data from the role-playing game was not recorded.

Looking over the visualization, it appears that I am not as active in conversations than my coworkers, which is true to some extent. However, this data fails to capture all areas of engagement and loses the context of the messages. For example:

• I may respond to a message containing a yes or no question by using the thumbs up or thumbs down emoji rather than replying yes or no.
• One of my coworkers is one of my housemates so we generally talk in person rather than online.
• This week we used the discord voice channels rather than the text channels for some of our conversations.
• Some messages were sent in the group chat that were directed towards specific people (rather than a direct message)
• I often send “block” or “paragraph” style messages rather than “sentence” style (i.e. one message may be equivalent to 2 or 3 from others)

Some trends can be identified from this data set:

• I exclusively send messages from the desktop discord application, but view on both my phone and desktop.
• The number of messages decline throughout the week with Tuesday having the greatest number of messages. This makes some sense as Tuesday’s are generally our busiest days of the week.
• Most messages are sent in the afternoon/evening. This also makes some sense because 75% of our labs run between 12:30 and 22:00.

It will be interesting to further collect this data to see if these trends hold.