After the fun and games of our tweetstorm, how has the redcorded data of the event stood up to our memories of what happened and what cnan it tell us about what took place?
Analysing the success of the event
Volume refers to the number of tweets in total and shows clearly that the average number of tweets rose dramatically from less than 50 a day to 187 on Friday. So from this data we can see that the tutorial had an influence in the frequency of tweeting on these days, we could even say that the tweetorial was a success. Or can we?
My interpretation of the statement found on the course website was that the tweet data we are analysing would be from the tweetorial. However, closer inspection shows that it is actually from a much longer period and was recording both before and after the tweetorial. This is key, I interpreted the instructions we were given in a specific way and the data I have been given access to does not quite fit the purpose I thought I had. This then also brings into question my findings that the tweetorial influenced the frequency of tweets. I can only say that the amount of tweets on Friday was different to other days in this period because I have been given data in a date range which included days on the outside of the tweetorial dates. So for the data I received, yes the data implies that there was indeed influence. However, I cannot see the same volume data for other days so it is therefore impossible for me to say if the tweetorial days showed higher tweet numbers than the rest of the course, if we had that data, we may indeed discover that the tweetorial days actually had a lower number of tweets than on other days and therefore don’t hold as much of an influence as we first thought. Therefore, from the data we have, we can only state that there were a certain amount of tweets recorded on a certain day. Not that the tweetorial did indeed have any kind of influence on tweet behaviours of the class nor that it was successful in creating meaningful discussion around certain topics.
Who was present and engaged?
Over the course of the recorded data, Phlip was the most prolific tweeter. This information is displayed as a league table which created a bit of competitiveness amongst the class about who was on the top ten and who wasn’t. One of my classmates mentioned a paper this week which discusses exactly this and says that displaying data in this fashion does indeed cause competitiveness (Cherry, T.L. & Ellis, L.V., 2005), I apologise I can’t remember who introduced me to this paper now. An interesting point to consider, however, is that there was no such competitiveness about who’s tweets showed the best understanding of the topic or the most learning, just about who was on the league of most prolific tweeters.
We could interpret this league table to say that Philip was engaged, or even the most engaged or that he was present and in learning analytics, this might be the way that this data would be used. We cannot say however that his tweets were of substance. He could have been retweeting the same message over and over in an attempt to influence the data rather than engaging with the debate or the tweets he was publishing may not have been engaging with the topics, they may just have been tweets which had the #mscedc hashtag and therefore were counted. Think roller skates and cheese.
Social learning
This leads me nicely to social learning, as Siemens (2013) points out “The learning process is essentially social and cannot be completely reduced to algorithms”, although we may interpret the high volume of tweets to say that discussions were taking place amongst peers on the topics given by Jeremy and James, the data doesn’t record the content of these tweets. It has however recorded a heat map of words used and as we would expect data and algorithm feature highly. However I’m drawn to the presence of other words, I’m, I’ve and perhaps. Does the use of these personal statements of I’ve or I’m show conversation and interpersonal discussion? Does it show students taking the learning and evaluating it in personal terms for their understanding? Does the high ranking of the word perhaps show uncertainty and lack of confidence in the topic? We can and do assume that the use of a social media platform encourages conversation and indeed social learning, but whow would we quantify this experience with data?
Accuracy of data
Although I am pretty sure, from my personal experience, that almost all of the tweets in this timeframe were in English, the data says otherwise. Therefore I must be wrong, I obviously don’t remember the events as well as I had assumed.
Not necessarily.
I know the 1 Swedish record was, in fact English, I know this because it was one of my tweets which twitter then offered to translate from Swedish when in fact it was in English. I also know that Colin and I both send tweets in Scottish Gaelic (Gah-lick not to be mistaken for Irish Gay-lick), after we saw this mistake, however as you can see from the chart, there is no record of Scottish Gaelic appearing, even though the key to languages says Twitter would recognise if there had been.
This emphasises nicely that we cannot guarantee the accuracy of data we are being given, which we then interpret to make judgements. There has clearly been an influence at work which has told the twitter archive algorithm that it should record languages a certain way and this hasn’t fit with what was going on. If my memory was infallible, I could say that there were only those 2 tweets which were not in English, therefore the data given by twitter is completely wrong. I cannot say that, however, but what I can say is that the data is definitely not accurate as I have detected at least one flaw, therefore we must question the data set as a whole. We do not know how the algorithm recorded this data to account for the flaws we have seen.
In conclusion. the problem is that we are being asked to analyse the data we have been given, meaning we should study it methodically to interpret its meaning and that’s the flash bulb, “interpret”. I have deliberately and repeatedly spoke of my interpretation of the data and the task. One person will interpret the information they have in one way and the next person may interpret it differently. External influences play a part on how we interpret data, the meaning or importance we place on things. What’s the phrase? Like looking at the world through rose-tinted spectacles, well mine are purple, so I’ll see it differently from you, but interpreting the data is only part of the story, once it has been interpreted, what is then to be done with those finding. For the sake of this example, if this information was being used as learning analytics, we could expect Philip’s tweet count to be associated with attendance or participation and in the same light if someone didn’t rank highly in the volume graph, would they then be marked lower for participation? As mentioned earlier, these stats don’t show the quality of the tweets, only their quantity, it is, therefore impossible to say that Philip was more engaged or participated in the discussion more than any other. We also cannot say who was present and participating but not actively tweeting.
I’m going to end this thought with repeating the word interpret. I have deliberately used this instead of analyse as there are associations based on these words, analyse, we associate with computing and therefore with accuracy. Interpret we see more as an art than a science and therefore holds the potential for human error, but as these few examples have shown, the computer can make mistakes and at the end of the day, it is a human who interprets/analysises the data and they can only work with the data they have been given to try to ascertain the information they need. Before acting on any analytical data, we should ask ourselves, how has this information been gained, why was it recorded and how was it recorded, before then interpreting the data for your purpose, all the while remembering what you have is interpretation, as Jeremy has said often over the last week, a proxy to help you interpret learning.
Refrences
Cherry, T.L. & Ellis, L.V., 2005. Does Rank-Order Grading Improve Student Performance? International Review of Economics Education, 4(1), pp.9–19. Available at: http://www.sciencedirect.com/science/article/pii/S1477388015301407.
Siemens, G. (2013) Learning Analytics: the emergence of a discipline. American Behavioral Scientist, 57(10): 1380-1400