There’s a very public debate about pipeline expansion unfolding in our own backyard (we call Vancouver, British Columbia home). We were curious to see what this contentious, divisive and globally significant issue would look like as a networked Twitter conversation.
We’re not picking sides, debating right versus wrong, or trying to separate fiction from fact. This case-study isn’t about content, sentiment or a rhetorical analysis.
We are focused on separating non-human from human social participation. This is a look at partisan programmatic amplification related to the Kinder Morgan pipeline expansion project.
To be 100% clear our position is that anonymous non-transparent, programmatic, bot-like online behavior is bad regardless of sides taken. We’ve looked at the “Red & Blue team” behavior in connection with the State of The Union Address (#SOTU). Our collaborative research helped produce an article about Programmatic Activism From Syria, which will be balanced by a forthcoming article in connection with Turkish computational propaganda and the hashtag #OperationOliveBranch.
Furthermore, this programmatic online behavior has an aggregate effect of distorting the reality of public participation. While the impact is difficult to measure, the use of bots (or other non-transparent marketing automation tools & techniques) shows a clear intent to manipulate public opinion or perception.
Finally, the automation of social feeds presents a significant threat to the overall information landscape. Giving “the machine” total or near total control of content distribution elevates the risk of exposure to cyber-threats associated with social engineering (hacked accounts) the delivery of malware infested content and phishing schemes.
We started this project with an intent to compare the two hashtags #StopKM and #SupportKM. Two hashtags, two rallies on Saturday, March 10, 2018, and two diametrically opposed positions.
The hashtag #SupportKM lacked enough data to make a reasonable comparison. We shifted our attention to the audience retweeting and mentioning the profile @fairquestions whose tweet appeared to be the impetus or at least the most engaged one in connection with the hashtag.
Our research highlights that both sides have supporters engaged in bot-like programmatic behaviors. (Data sets and methods located in concluding appendix below).
#StopKM Participant Findings
We collected 832 unique profiles from 2400 tweets. These profiles tweeted at least once using #StopKM (regardless of side). From this group we identified 147 partisan profiles (those against the pipeline expansion project) that had tweeted at least twice.
We classified 49 of these 147 profile as “Cyborgs” having tweeted 50 + times per day on average over the last 7 days. We further segmented these 49 “Cyborgs” —
- 400+ tweets per day = 2
- 300+ tweets per day = 2
- 200+ tweets per day = 3
- 100+ tweets per day = 23
- 50–99 tweets per day = 19
- Total anonymous “cyborgs”= 34
- Real (non-verified identity) “cyborgs” = 13
- Cause or organisations tweeting at a “cyborg” rate= 2
- Total anonymous profiles from multiple tweet data-set = 79 out of 147
The Top Four “Cyborgs”
Audience Findings from @fairquestions
We collected 402 unique profiles from 1400 tweets. These profiles had mentioned or retweeted @fairquestions at least once (regardless of side). From this smaller data set we identified 158 partisan profiles (those aligned with the pro-industry position of @fairquestions) that had tweeted at least twice.
We classified 44 of these 158 profile as “Cyborgs” having tweeted 50 + times per day on average over the last 7 days. We further segmented these 44 “Cyborgs” —
- 400+ tweets per day = 1
- 300+ tweets per day = 1
- 200+ tweets per day = 4
- 100+ tweets per day = 14
- 50–99 tweets per day = 24
- Total anonymous “cyborgs” = 30
- Real (non-verified identity) “cyborgs” = 12
- Cause or organisations tweeting at a “cyborg” rate = 2
- Total anonymous profiles from multiple tweet data-set = 103 out of 158
The Top Four “Cyborgs”
Moving from a snapshot of the players behaving in un-human fashion, we compare how the respective hashtags look as networked conversations.
Putting these maps in context:
- Both moments from March 11th, at 1603 (PST) for each hashtag (previous 200 tweets)
- These are the 50 profiles that tweeted each respective hashtag the most.
- We have not differentiated the profile on these maps based partisanship.
- Thickness of the lines (edges) between profiles (nodes) not directly connected to the hashtag represents the number of tweets (mentions of that profile).
- Secondary hashtags are the most used in conjunction with the central hashtag.