Twitter is a public, social media platform which, according to their own statistics, has 313 million monthly active users. Users are able to post ‘tweets’ of up to 140 characters as well as sharing pictures, videos or external links. As one of the so-called ‘Big 4’ social media platforms, Twitter is one of the biggest mediums through which people are able to voice their opinions as well as read and absorb those of the people and accounts they follow. AS the pioneers of the hashtag function, Twitter conversations can often be dominated by a specific topic for a particular period of time depending on the popular/trending hashtags.
Of all the social media platforms, Twitter draws one of the most diverse audiences spanning age, gender, political inclination and a broad mix of individuals and brands. It also has the advantage of conversations which are free from a lot of the noise that is inevitably found in conversations on on other platforms such as Facebook. It is unclear whether this is indicative of the type of people that use Twitter or a result of the system itself, limiting the amount users are able to say which in turn makes every word more precious; but either way, it is a unique advantage that Twitter holds against other social media platforms.
The automated task I set up on twitter was quite basic however it was a set of results I was interested in recording and comparing. I first set up a search rule to record every time the exact phrase “women should” was used in a tweet written in English before repeating the rule with the phrase “men should.” From a quick scan of the results I realised that they were unfairly skewed by references to a quote/quotes from Phyllis Schlafly, an infamous conservative activist, who had died on September 5. I adjusted my search parameters to remove any tweets which contained the words Phyllis or Schlafly as well as those which contained the hashtag #PhyllisSchlafly and plotted the result on the bar graph below:
The results confirmed what I had expected before conducting the process however it is interesting to see just how starkly contrasting the two figures are, particularly when visualised as in the above graph. The obvious flaw in my automated process, as I have conducted, it is the short time period of only 5 days. Ideally I would have liked to conduct the process over a period of at least 2 weeks to get an accurate gauge however the methods and applications I was using only recorded as far back as September 2. In saying that, the sample space I ended up with was still comfortably in the thousands as it is obviously a commonly used phrase amongst twitter users, so I still feel I was able to gain a somewhat accurate insight into the broader trends.
Following on from this initial, simple automated process, I decided to look more closely at the tweets containing “women should.” This second stage involved taking a sample of the words or phrases that followed “women should…” in the previous tweets to get a more in-depth sense of what twitter users think. Partly inspired by the example regarding ‘smells like’ in the lecture by Chris Gaul, I took a random selection of 20 of the collected tweets and recorded how they completed the sentence “women should…”
It is clear after this exercise that there is another major flaw in my automated process, particularly if I wish to use it for more than simply comparative purposes. The simplicity of the search parameters and filtering in my initial probe means that it ignores factors such as sarcasm or satire and does not take into account the general context of the statement so in actual fact the tweet could read “women should not…” or perhaps be a retweet of someone else’s statement which the person retweeting disagrees with entirely in their own part of the tweet. Determining context or intention in a written medium can be complex even for human readers so it is intrinsically tricky to program into a robot and is something that is likely beyond my current skill level however if I were to repeat this task in the future, I would ensure that I was able to find some way to achieve a set of results that more accurately reflected what I was searching for and allowed me to work more with the resulting data.
In terms of a visual design response that could be generated from the data/results, I think it could be interesting to move away from the comparison between men v women graphics that tend to, unfortunately, make a large portion of the audience switch off. Instead, I think an interesting visual response could be to visualise, possibly illustrate, some of the more ridiculous things that people on twitter believe women ‘should’ do as a means of highlighting how we as a society need to stop thinking we have the right to tell anyone what they ought to be doing with their life, regardless of gender.