Canadian Pipeline Politics: Mapping (visual) discourse in platform spaces

Team Members

Carrie Karsgaard, University of Alberta
Lucia Bainotti, University of Milan
Serena Del Nero, Politecnico di Milano, DensityDesign
Giacomo Flaim, Politecnico di Milano, DensityDesign
Michael Hockenhull, IT-University of Copenhagen
Maggie MacDonald , Concordia University
Antonio Martella, University of Pisa
Erika Valderrama, London School of Economics
Gabriel Valerio, Tecnologico de Monterrey

Contents

  1. Summary of Key Findings
  2. Introduction
  3. Initial Data Sets
  4. Research Questions
  5. Methodology
  6. Findings and Discussion
  7. Conclusion
  8. References

1. Summary of Key Findings

Methodological findings: Twitter

Twitter’s free-form user location field enables political expression where location-identifiers have political significance; in the Canadian landscape, for instance, users may identify according to Indigenous place names rather than officially sanctioned place names.

The implication for digital methods research is that user location provides a means of exploring issue alignment and political stance for issues with geographical components.

Methodological findings: Instagram

An analytical challenge - but also an opportunity - is posed by Instagram’s multiple discursive spaces - images, text, and hashtags - as they are variously used over time. The project prototypes a comparative approach to these multiple spaces through analysis of each through multiple time slices, including analysis of the linkages between various discourses where possible (i.e. by hashtag-image and text-image analysis). Included is a new approach to text-image analysis using Cortext, which allows exploration of the user-generated text in Instagram posts, beyond what is available via hashtag analysis.

Taken together, multiple maps reinforce certain issue patterns through their repeated representation in various visualizations; at the same time, individual maps reveal nuances of the issue that only emerge through a single discourse (whether visual, textual, or connective via hashtags). This project thus demonstrates how critical discourse analysis and visual analysis may be conducted at multiple and intersecting levels through a critical cartographical approach, enabling a more robust understanding of the issue as it is performed online.

Substantive Findings

Our findings indicate that when analyzed as above, the tools embedded in both Twitter and Instagram allow us to infer discursive alignment with issue positioning, not only for/against the key issue, but also within sub-groups, allowing a nuanced view of the issue. For instance, within anti-pipeline sentiment, analysis of locations, hashtags, text, and images reveals competing ideals between protection of the land (as pristine), ownership of the land (as a Vancouver resident), and stewardship of the land (as already occupied by Indigenous peoples). By tracing these discursive groups over time, we see increasing overlap within our issue network visualizations, where distinct clusters are replaced by heterogenous networks, indicating that the pipeline issue may function as a boundary object, bringing various publics closer together.

Presentation

Our presentation can be found HERE.

2. Introduction

In Canada, known to many Indigenous peoples as Turtle Island, the Trans Mountain pipeline controversy reflects conflicting and hierarchical worldviews and agendas from a multiplicity of stakeholders including the environmental sector, Indigenous activists, governments (civic, provincial, and national), corporations, and interested individuals. This project explores how, within Canada’s colonial context, hierarchies of actors and discourses about the Trans Mountain pipeline are arranged, signifying epistemological divides between these actors, including even mainstream environmentalists and Indigenous protectors of the earth. This analysis is important, for there is currently great incongruity between the increasing recognition of Indigenous peoples’ knowledges and concerns with regards to land, environment, and climate issues (Green & Raygorodetsky, 2010; Tsosie, 2007) and ongoing colonial violences linked with extractive practices. Though Canada is striving for reconciliation with Indigenous peoples, it is difficult to imagine true reconciliation and justice without opening up a conversation about the colonial underpinnings of current pipeline dialogue.

The Trans Mountain pipeline controversy is particularly fitting for this study as it has engaged various publics in both active support and resistance. As an overview, Canada’s Prime Minister, Justin Trudeau, has approved this pipeline expansion from Edmonton, Alberta to Vancouver, British Columbia (BC), to be built by energy company Kinder Morgan. Supporters, including Alberta’s provincial government, argue that the pipeline is necessary to Canada’s economy via oil exports, and that it will provide jobs for countless Albertans. By contrast, some environmentalists and politicians in BC oppose the project, warning of the pipeline’s environmental hazards and contributions to climate change. Indigenous communities are similarly concerned with environmental destruction - but also with Canada’s lack of consultation with them as sovereign peoples on land that has never been ceded to the Canadian government. Resistance has taken the form of everything from lawsuits to protests and demonstrations, and the controversy has led to trade wars between the provinces of Alberta and British Columbia. Supporters have called Trudeau to invoke the Federal Emergencies Act in order to halt the lawsuits and begin construction; resistors have drawn Trudeau’s attention to the UN Declaration on the Rights of Indigenous Peoples (UNDRIP) and his legal requirement to receive “free, prior, and informed consent” from Indigenous peoples before acting. The conflict runs deep, and the issue gets at the heart of Canada’s nature as a settler colonial state (Tuck, McKenzie, & McCoy, 2014), wherein settler centrality and superiority is naturalized through policy, law, ideology, and culture, at the expense of Indigenous peoples who continue to be displaced from the land, which is conceptualized as a “resource.”

It is likely that a settler-colonial approach to resource extraction can easily be seen in the discourse of pipeline supporters, but its persistence within the environmental sector is more subtle. Environmentalists, of course, advocate against resource extraction for such reasons as the preservation of wild land and prevention of climate change. In Canadian settler colonialism, however, the “environment” is habituated in relation to Canadian national identity through romantic ideologies that characterize environmental thought through a nature/culture divide, wherein human actors dominate and manage nature through means as diverse as conquest and conservation (see Cronon, 1996; Gomez-Pompa, 1992). Problematically, “environmentalists have found it easier to advocate protection of ‘natural’ environments and warm furry animals than to prioritise protection of the rights of indigenous peoples whose stewardship of habitats and use of many warm furry animals is harder to encapsulate as a bumper sticker. Environmentalists have often opposed indigenous use and occupation of (even access to) lands they classify as having high conservation values” (Howitt, 2001, p. 26).

Resistance to the Trans Mountain pipeline controversy, however, has in some cases brought together Indigenous groups and environmentalists, potentially functioning as a “boundary object” (Star and Griesemer, 1989) between groups divided by settler colonialism. With the large-scale public expressions available on Instagram and Twitter, this project aims to map the extent to which epistemological differences are reflected in the more subtle issue agendas of various groups, despite positional alignment for/against the Trans Mountain pipeline - as well as the extent to which the pipeline controversy is bringing groups closer together.

Exploring this issue on Instagram and Twitter also enables the development of multiple network counter-maps, in order to “[unsettle] the very categories that constitute the intelligibility of modern power relations” (Crampton, 2010, p. 125) in Canada’s colonial context. Through both semantic analysis and visual methodologies, this project traces the alignment, divergence, and hierarchies of various issue publics as expressed both textually and visually. In doing so, it engages with the multiple grammars available through Twitter and Instagram, including location tagging, hashtagging, textual expression, and visual expression, to explore how platform dynamics allow (or disallow) various means of expressing issue alignment.

3. Initial Data Sets

Both Instagram and Twitter were queried for the following terms, which include pro-pipeline, anti-pipeline, and neutral terms:

stopkm
buildkm
kindermorgan
stopthekmbailout
stopkindermorgan
notmx
transmountain
transmountainpipeline
protecttheinlet
tinyhousewarriors
keepcanadaworking

3.1 Twitter

Using the DMI TCAT (Borra; Rieder, 2014), tweets were extracted in a query bin from June 26, 2018 to July 7, 2018, totalling 27,000 tweets from 13,000 users. While this is a relatively short time period and small data set, the July 3 blockading of oil tankers by climbers suspended from a Vancouver bridge made this period interesting for our issue as it likely provides a spike in Twitter activity.

3.2 Instagram

A set of all Instagram posts linked with the query terms above was extracted using the “--tag” function on instagram-scraper, resulting in a data set of 12,963 posts from 07/30/11 - 07/06/18. This large data set was reduced by creating five (5) key time slices according to spikes in Instagram activity, which correspond with significant events in the controversy, as mapped against Google Trends. Each slice is characterized by a 3-week period, defined by one week prior to the spike and one week following, as per the chart below.

4. Research Questions

As expressed on Twitter and Instagram, to what extent are actors and discourses about the Trans Mountain pipeline arranged along a colonial divide, signifying epistemological divides between these actors, and how do these arrangements map onto sentiments for/against the key issue?

  • How can the modes of each platform, including (a) the grammars of hashtags, (b) textual content, (c) patterns of visual imagery, and (d) user profile location, allow us to infer discursive alignment within the adopted positioning, not only for/against the key issue, but also within sub-groups of support or resistance?

  • What are the common visual and textual discourses identified, and to what extent are they colonially aligned?

  • How do Instagram and Twitter differ in the above?

5. Methodology

This mixed-methods project follows a social cartography approach to issue mapping (Marres & Moats, 2015; Marres & Weltevrede, 2013; Rogers, Sánchez-Querubín, & Kil, 2015) using both Instagram and Twitter data. Through both semantic analysis and visual methodologies, we trace the alignment, divergence, and hierarchies of various issue publics as expressed both texually and visually. Our work with developing the social cartographies and performing analysis is iterative, hermeneutic and data-focused in line with the data sprint approach (Venturini et al, 2018).

5.1 Twitter

Twitter users can choose to identify their profile with a location that is self-determined. This means that alongside officially sanctioned settler-colonial named locations (ie: Edmonton, Vancouver) there are also users self identifying according to Indigenous place names (ie: Turtle Island, Treaty 6 Territory, Unceded Coast Salish Territory), suggesting a user base that actively identifies with Indigenous sovereignty and solidarity.

Drawing from the DMI TCAT data set, we used OpenRefine to categorically sort locations under three identifiers (location1=city, or indigenous territory, location2 = province, or the signifier ‘Indigenous’, location3 = country). From the TCAT data set of 27,000 tweets and 13,000 users, a total of 5098 users produced 12,416 tweets from various locations in Canada. Of that data, we identified areas adjacent to the boundary object, or those most near to the development of the pipeline, namely within Alberta (AB), British Columbia (BC) and the various Indigenous Territories (IN).

  1. Network visualization in Gephi using the pre-extracted TCAT dataset.

    1. A co-hashtag graph
    2. a user-hashtag graph
    3. a user-media graph
  2. An Excel Macro script built to consolidate tweets under their user and cluster them by user-identified locations.

  3. Tableau for stats and visualisations of the dataset, which is characterised by the following traits:

    1. 27.000 tweets
    2. 13.000 users
    3. Posted over 15 days, from 26/06 to 10/07
    4. The most active user had 286 tweets
    5. 3.7% of tweets contain media files such as images, gifs or video
    6. 22% of tweets contain links to other twitter posts or external links

Using the user-identified profile locations, we consolidated the locations into recognizable approximations according to city, province, and country. This consolidation was done by using first an Excel macro that consolidated the tweets under their user, which significantly reduced labour time by allowing us to manually tag location per user rather than per tweet and re-link the data back to the tweets afterwards.

Once this consolidation was complete, we emphasized the category country “Canada,” and three key locations within it, namely “BC” (British Columbia), “AB” (Alberta), and “IN” (Indigenous identified), to isolate the most relevant groups for this research topic. We also categorized all other provinces under a general Canada cluster.

Once the data set was cleaned, we used Table2net (http://tools.medialab.sciences-po.fr/table2net/) to convert the table to a Gephi-friendly format and mapped the correlations between tweets and locations for the full dataset in Canada. This mapping was done according to user/hashtag and then location/hashtag. The user/hashtag connection revealed a few users who flooded their tweets with relevant activist hashtagging but used minimal text outside of that. We investigated these users to see if they were bots (www.botometer.com; www.botcheck.me), and it appears none of them were; they were simply very active anti-pipeline protesters, and almost without exception they identified as Indigenous or in solidarity with Indigenous territories.

To further understand the ideological discourses embedded in the hashtags, we developed an excel macro script. This script obtains all the hashtags contained in a tweet and linked to the user’s location in a matrix. This matrix also was automatically converted into a list which identifies the sharing hashtag among locations and the unique ones. This list was imported into the already existing co-hashtag network as metadata, and could then be used to partition the hashtags in the graph based on the location of the users tweeting it. By maintaining the same graph and partitioning it differently, it was possible to compare the original modularity and thematic-based analysis of the hashtag clusters with the locations data, adding a layer of analysis not present in the “raw” data from the Twitter platform. The majority of the hashtags were of course shared by two or more groups of users, but it was also clear that many hashtags were unique to particular user groups, as will be elaborated on below.

5.2 Instagram

Instagram methods focused only on slices 1, 3, and 5, in order to narrow our data set; these are contextualized against grounded events in the image below. We followed multiple methods in order to compare what (a) the grammars of hashtags, (b) textual content, and (c) patterns of visual imagery allow us to infer of the discursive alignment within the adopted positioning.

As location metadata is unavailable through instagram-scraper, we could not create a comparative case for Twitter based on location. However, such a comparison would not in any case be possible as Instagram locations are pre-set (not free-form), disallowing the forms of location expression allowed by Twitter.

  1. Co-hashtag network in Gephi, discluding search query terms. Allows for network and cluster analysis based on hashtag usage.

    1. WordCloud: based on co-hashtag modularity from Slice 1, WordClouds provide discursive analysis of issue clusters based on hashtag frequency. Minimum settings vary depending on the sample size.

  2. Bipartite network (hashtag-images) in Gephi, discluding search query terms and discluding hashtags with <10 occurrences. Allows for network and cluster analysis based on hashtag and image usage together.

  3. Cortext multi-step process.

    1. Terms extraction for top 100 terms (2-3 words; minimum frequency of 3).
    2. Co-term network.
    3. Gephi bipartite network (term-images). Allows for network and cluster analysis based on term and image usage together.
  4. Memespector and Google Vision API. Gephi bipartite network (image-label) allows for network and cluster analysis based on imagery patterns.

6. Findings and Discussion

6.1 Twitter

General visualizations and stats

6.1.1 Co Hashtag Network Analysis

From the TCAT dataset, a co-hashtag graph identified three main clusters, an extended “halo” cluster and some satellite clusters.

  1. A pro-pipeline cluster
  2. An anti-pipeline cluster with climate orientation
  3. An anti-pipeline cluster with Indigenous orientation
  4. A spread out “halo” cluster of international location hashtags and climate-related hashtags
  5. A satellite cluster with Ontario-local hashtags.

These clusters seemed to correspond well to the modularity-algorithm’s suggestion for clusters and colour-schemes.

6.1.1.1 User-hashtag analysis

The user-hashtag graph is dominated by large central hashtags, with many shared users between them and few but significant sub-clusters on the periphery of the graph. As could be expected, the dominant hashtags are neutral (#kindermorgan, #transmountain), regional (#cdnpoli, #bcpoli) or against the pipeline (#stopkm). One periphery cluster is pro-pipeline, with multiple users arranged around the hashtag #keepcanadaworking, yet away from the centre of the graph. This graph, while interesting, seems to contain the same information found in the co-hashtag graph (described below), just presented in a different way. As our topic is more focused on discursive and epistemological aspects of the dataset rather than on user relations, we decided to pursue the refinement of the co-hashtag graph instead of developing the user-hashtag one.

6.1.1.2 User-media analysis

In the user-media graph, users center around major media outlets. However, media outlets seem not to be clustered according to political lines, and instead are mashed in a central cluster with twitter links (replies to other tweets) forming the heart of the cluster. This is perhaps unsurprising as the Twitter dataset was collected during an event (blockading of oil tankers) covered by many news outlets, where much activity by Twitter users may consist in simply retweeting the newest update on the event itself. A close reading of the content of the tweets might help problematise this hypothesis or reveal political orientation (i.e. critique or support of various media content); however, this is beyond the scope of our study.

6.1.1.3 Co-hashtag analysis

The co-hashtag network graph (below) reveals that pipeline-resistant sentiment contains several subgroups. Indigenous and climate issues (#waterprotectors, #decolonize, #climateaction) appear together as a distinct group from Canadian pipeline politics (#canpoli, #Trudeau, #liberal). A distinct regional Ontario cluster can be identified naming cities across the province linked to such hashtags as #onpoli, #toronto, #nokmbuyout and #stopkmbuyout; further research could perhaps reveal whether this is an anti-pipeline solidarity cluster or whether it involves references to Canada’s national government as located in Ottawa, Ontario.

Interestingly, included in the dispersed purple cluster is a collection of international locations linked with hashtags referencing climate change and fossil fuel divestment, linking the Canadian issue with global climate and oil politics more broadly. This map is the only location in our study where we see the emergence of this theme and the global positioning of the Trans Mountain controversy.



6.1.2 Locations as ideological discourse

Free-form user location on the platform creates space for users to oppose colonial mapping, or, to actively identify themselves in an Indigenous or anti-colonial way. The user-identified place of origin may reveal how the user identifies in relation to the settler-colonial state. For example, users may use Indigenous place names, such as the Cree “Amiskwaciwâskahikan,” in place of “official” city names like Edmonton. Or, they may deny the settler-colonial state by replacing the name “Canada” with “Turtle Island,” the Indigenous name for North America. Finally, they may draw attention to Canada’s history of treaty-making with Indigenous peoples and theft of Indigenous lands by referencing such places as “Treaty 6 Territory” or “Unceded Syilx Territory.”

We found empirically that these locations map onto the hashtag-based discursive space in a way that is consistent with the content of the discourses, suggesting that users on Twitter do indeed engage in toponymic politics (Rogers, 2015). However mapping the two onto one another also produces some surprising results, such as the cluster of ‘Ontario locations’ and ‘halo’ of ‘Canada/International geographical locations…’ identifying toponymically as Indigenous. Further research would be needed to determine the reason for this.

In co-hashtag analysis, Indigenous users differ significantly in their terms from British Columbia users, despite both expressing anti-pipeline sentiment. British Columbia-based tweets center around terms of ownership of the environment, preservation of wildlife, and a NIMBY (not in my back yard) approach to pipeline protest. Indigenous terms tend towards stewardship rather than ownership of the land as well as highlighting protest action, with hashtags like #waterprotectors, #peacefulprotest, #tinyhousewarriors and #peoplenotprofits appearing in Indigenous users’ tweets.

In general, it should be kept in mind that these findings indicate that users engage in toponymical politics on Twitter, but that the extent and character of these politics need further examination, preferably through qualitative and close-reading approaches.

6.2 Instagram

6.2.1 Beginning with hashtags: promoted issue affiliations

6.2.1.1 Co-hashtag analysis

A co-hashtag analysis establishes an initial overview of the different discourses around the Kinder Morgan pipeline issue over time.

Initially, the co-hashtag analysis provided us with very dense and interconnected networks, in which it was hard to identify different clusters and extract the different issues at stake. This was the case, firstly, because all the hashtags were very much connected with the query terms, thus creating a dense hub around them. Secondly, many hashtags were related to each other, reflecting a platform-specific tagging behaviour. Indeed, the majority of posts contained a large amount of hashtags that were recurrent and used to contextualize the posts themselves within the pipeline controversy. When removing the query terms from the analysis, it was possible to see more clearly some of the different discourses.*

The images and chart below provide a comparison of the top five modularity classes over time, according to Slices 1, 3 and 5. A number of themes emerge in response to our research questions:

  • Pro-pipeline sentiment is less evident in the co-hashtag analysis. This is likely due to the fact that the pipeline is already officially sanctioned, so public discussion is focused on resistance. It may also speak to issues of self-selection in terms of the use base of a platform such as Instagram, which may be populated mainly by actors with anti-pipeline sentiment.

  • Anti-pipeline sentiment takes a variety of forms, reflecting some of the colonial differences identified in the introduction above:

    • The proportion of explicitly Indigenous issues shift over time but are present in all three slices. Indigenous-related hashtags include (a) issues of consent and sovereignty, including references to UNDRIP, (b) depictions of the land as sacred and connected to life, (c) connections to the occupation of Standing Rock and #nodapl resistance in the United States

    • A form of environmental conservation particularly around ocean life is present in each slice, though hashtag usage shifts the discourse somewhat between the three time periods. A key difference appears in Slice 3, which leverages World Water Day for conservation messaging, in contrast with the more generic “save the orcas” and “save our oceans” messaging of the other two slices. These conservation clusters do not echo an Indigenous notion that “water is life,” but they instead depict the natural world as something to be protected in and of itself, reflecting a more colonial preservationist orientation to the land.

    • Both local pride and discourses around travel in BC more generally and Vancouver more specifically indicate a relation to land as something to be protected for human enjoyment, again reflecting a more colonial orientation to the land as commodity.

  • A blend of civic (i.e. Vancouver, Burnaby), provincial (Alberta, BC), and national politics emerge in relation to the pipeline, indicating jurisdictional issues.

*It is standard methodological practice in digital methods and network analysis to “normalise“ a dataset when one knows the query term that the corpus is constructed on the basis of. This is done based on the reasoning that the query term will be connected to all other queries, and thus will overshadow other terms or structures of the network if not removed. In this particular instance normalising is more problematic as the query is constructed from multiple queries, and by deleting them we also inadvertently delete “natural” occurrences of the terms. However it proved to be a pragmatically advantageous move which allowed existing clusters in the network to appear more readily.



Slice 1

Slice 3

Slice 5

25.28% - Conservation and BC quality of life


ocean, salishsea, orcas, rawvegan, healthyBC, protest, tourism, lovewhereyoulive

11.74% - World Water Day 2018


waterislife, waterdefenders, stoppipelines, pinotnotpipelines, protectourcoast, pnw, saveoursalmon

16.31% - Canadian and provincial pipeline politics


canada, pipeline, trudeau


locations (city, province names), key actors (political parties, people)

11.36% - Broad environmental news and government


climatechange, electoralreform, media, JustinTrudeau, parisagreement, greenenergy, solar, fuckoil

10.19% - Generic nature and environmental travel


naturelover, protectourplanet, spiritual, mothernaturerocks, travel, texas, mountains, camping

11.17% - Anti-pipeline sentiment and protest, including Indigenous and general activist terms


keepitintheground, renewables, nopipelines, decolonialmemequeens, waterislife, fuckcapitalism, respectmypronouns, affordablehousing, metoo

9.33% - Anti-pipeline with Indigenous focus

noconsent, waterissacred, standingrock, nodapl, uncededterritory, redefinewealth, solidarity, mothernature

9.22% - Canadian and provincial pipeline politics


falsedeau, undrip, vanpoli, townhall, tmx, albertaoil, oilsandsproud, meanwhileinalberta, transcanada, parisagreement

8.87% - Anti pipeline with a BC/Vancouver/water focus


BeautifulBC, saveourocean, nospills, dirtyoil, BCstrong

8.17% - Pipeline support


fucktards, workingclass, laypipe, bluecollar, oilandgas, veterans, consultation, nothappy, theregoestheneighborhood

9.17% - Indigenous presence on Burnaby Mountain and Indigenous rights


pipeline, burnabymountian, notankers, colonialism, noconsent, rcmp, bitumendontkillmyvibe, riseup, nojurisdiction, supportindigenoussovereignty

6.24% - Identity crisis! Half sustainability / half pro-pipeline


organicfood, animals, volunteer, restoration, international, data


interns, trades, design, Ilovepipelines, pickuptruck, laythepipe, albertaoil

6.81% - Vancouver/BC specific, pride of place


squamish, walks, kitimat, export, vanlife, children, health, wellbeing, solarpower, epicviews

8.44% - World Water day 2.0 and conservation


canadaforsale, greenpeace, civildisobedience, notinmybackyard, savetheinlet, ecojustice, protectwater, treehugging, renewable

6.1% - Travel, adventure


destinations/locations (USA, mexico, mountains, etc.), family, running, cycling, photography, trains, traintravel

To further explore issue alignment, we visualized hashtag frequency within key modularity clusters in Slice 1 using Word Cloud, as below. Taken collectively, hashtags reveal colonial differences among anti-pipeline sub-groups. For instance, the first cluster is dominated by hashtags that link the Trans Mountain controversy to Standing Rock and the Dakota Access Pipeline in the United States, referencing Indigenous rights and the need for Indigenous consent to projects carried out on Indigenous land. In this cluster, a sense of Indigenous stewardship over and rights to the land are strongly asserted. By contrast, while the next cluster references rights, it is dominated by Vancouver-centric language and city pride, indicating a sense of ownership over the land by residents who consider “#vancouverisawesome,” in a more colonial understanding of land as property. The third cluster detailed below focuses on ocean health and protection of the orcas, in a more traditional environmentalist approach to the land as pristine and empty, to be protected from human impact. While these alignments warrant more detailed analysis and comparison over time (according to the three Slices), it is clear that a co-hashtag network may indeed reveal differences in issue alignment and epistemology even within a shared resistant space.



6.2.1.1 Hashtag-image analysis

Following the image-centric nature of Instagram, we next traced the visual discourses that emerged in relation to the hashtags by creating image-hashtag networks.

In Slice 1 (below), visual discourses emerge in relation to the textual ones outlined above. For instance, around the “nodapl” hashtag are images of protests - and also of Indigenous imagery, “water is life” placards and posters, visual references to #coastprotectors and #waterprotectors, and images surrounded with a #noconsent frame, all of which reinforce Indigenous rights to consultation and connectedness to land. Again as above, a Vancouver-centric cluster emerges where images of Vancouver landscapes and architecture are interspersed with protest photos, demonstrating love for the city and perhaps a NIMBY (“not in my back yard”) attitude. Within the “save the orcas” cluster are repeated images of whales in the sunset, visually mirroring the hashtag discourse of pristine oceanic landscapes, to be protected from oil spills. Interestingly, a pro-pipeline cluster is visible, aligned with hashtags such as #canadaproud and #oilsandsproud, which is dominated by images of white men in suits or collared shirts alongside pro-pipeline quotes, as well as infographics connecting pipelines to progress by Canada Action, a pro-pipeline movement. With their appeals to traditional, gendered authority, these images contrast starkly with resistant images.

Slices 3 and 5 diverge from the co-hashtag patterns above, as well as the Slice 1 image-hashtag network, insofar as they do not present clearly identifiable clusters. Slice 3 is dominated by water-related hashtags, interspersed with a mix of Indigenous imagery, protest photos, and environmental imagery of both land and water. It is possible here that World Water Day is dominating the hashtag space somewhat, linking all of the related images together. Noticeable for the first time, however, is a collection of images of police in the centre of the network; we can here see the intensification of the controversy, at least from the perspective of the state and law enforcement.

Slice 5 is similarly constituted by a heterogeneous collection of hashtags and photos, though this collection appears to be constituted of more memes, illustrations, and reposted images than previous slices. Such recycling of material could be attributed to alignment of this spike with Canada’s purchase of the pipeline rather than with a specific protest, as people post existing images in protest to this purchase rather than sharing live images from an on-site event.

Furthermore, Slice 5 is characterized by a collection of strongly worded hashtags related to colonization, including #decolonizeyourthinking, #stopcolonialviolence, #indigenousrights, #justintrudeauisacolonizer and #canadasupportscolonization, likely in a strong anti-colonial response to Trudeau’s pipeline purchase.

6.2.1.2 Beginning with text: expressed (textual) issue affiliations

Textual analysis via Cortext allows exploration of the user-generated text in Instagram posts, beyond what is available through hashtag analysis. This is key as Instagram does not limit text (as Twitter does), and users may express more detailed and personalized content beyond what they express via hashtags, which are typically used to connect the post to the larger issue community or for promotional purposes.* By extracting key terms and then creating networks with these terms as they are used in Instagram posts alongside images, we hoped to explore more deeply various issue alignments than is possible through hashtag analysis. This key term extraction procedure was a preliminary or prototype attempt in our research. Subsequent analysis of the key term-image network required extra interpretative efforts, since the terms extracted and their relations to the images were not always straightforward. We believe that the networks produced would be more legible, if the process of extracting terms was refined further. As it stands, we do however believe that analysis of the graphs has provided a valuable additional perspective on Instagram as an issue space in this controversy.

In Slice 1 (below), similar visual and textual discourses emerge to what is also visible through hashtag analysis, particularly in relation to (a) pipeline protests and (b) ocean protection. What is most interesting about this network is a connection between Indigenous peoples and climate issues. Further, it is through climate discourse that Indigenous peoples are linked with Canadian Prime Minister Justin Trudeau; this is a slightly different (yet related) link from the issue of consent and UNDRIP, which emerges in the hashtag graphs above. This difference warrants further exploration, perhaps by looking closely at individual tweets or through comparisons with other network graphs.

*It should be noted, that whilst Instagram allows for as much text to accompany an image as users want, there is no minimum amount, meaning that some users may post very little text at all. This risks skewing the key term extraction and any attempts at semantic analysis, which therefore should be undertaken with care.

Slice 3 similarly reflects a connection between climate issues and Indigenous peoples, again in relation to political leaders; this connection clearly warrants further exploration. Interestingly, this network reveals two collections related to water: (a) visual water discourse, focusing on the protection of ocean life (particularly whales) in relation to oil spills, and (b) textual water discourse, referencing water protectors and “water is life,” in conjunction with visual imagery of protests and banners. These dual water discourses indicate different understandings of land, signifying underlying ontological differences within the anti-pipeline camp. Interestingly, a separate “protect the inlet” cluster stands apart from both of these; Protect the Inlet is an anti-pipeline campaign with strong ties to Indigenous communities, including the Kwekwecnewtxw Watch House, which constitutes a core Indigenous and allied space for resistance to the pipeline. Though this cluster is not directly linked with the other water-related clusters, its imagery more closely mirrors that of the water protectors cluster (b) through photos of protests and banners.

Slice 5 is markedly different from the previous two slices through its discursive emphasis on the pipeline itself, as well as the emergence of a cluster relating to Trudeau. From this network it is clear that Instagram users are posting about the government’s purchase of the pipeline, centring their discussion on the Prime Minister. As noted in the image-hashtag network for Slice 5 above, this network is dominated by memes and reposted images, and Trudeau is featured in a number of these. Further research into the images and memes of Trudeau would likely reveal the specific critiques that emerge of the Prime Minister at this point in the controversy.

6.2.1.3 Beginning with images: expressed (visual) issue affiliations

To build image-label networks, we used the python version of Bernhard Rieder memespector script made by André Mintz (2018) and available on https://github.com/amintz/memespector-python. Due to some downloading problems that halted the script many times in an unintelligible way, we had to collect all images in order to build image-label networks. Because of lack of time, we were not able to connect images to original metadata; this is unfortunate, as the metadata would be useful in helping us see whether these visual patterns reflect other textual or user patterns, providing deeper insight into the issue. An additional issue is that we had to cut less frequent images due to the size of the network, which means the smaller pro-pipeline collection does not appear in the network, reducing our categories of analysis.

At the same time, however, these networks do allow us to see the dominant visual discourses as well as differences over time.

Slice 1 is dominated by protests (and close-up images of people in protests), as well as natural landscapes, with a small section of posters/cards/illustrations. This imagery reflects anti-pipeline sentiment as resistant, grassroots, and urban, particularly as the protests are pictured in urban landscapes. At the same time, it also shows the connection of the pipeline controversy to concerns of environment and landscape, whether ocean or mountainous landscapes.

Slice 3 is similarly dominated by protests, but here the environment (above all forests and trees) occupies the central cluster of the network, in contrast with the urban protest of Slice 1. This draws attention to the located nature of protests, particularly for an issue like this which is strongly geographical in nature. Protests occur in locations that are key to the issue, including urban and rural spaces, as well as in ocean landscapes.

As with Slices 1 & 3, Slice 5 indicates another protest location, with a large cluster of “kayactivists” who have taken to the ocean. This is the first slice that features repeated images of Alberta Prime Minister Rachel Notley - here, hugging her dog, in what is potentially a public relations campaign. Finally, there emerges a larger cluster of vehicles than in previous slices; whether these are connected to pro-pipeline sentiment warrants further investigation into the actual Instagram posts.

Through this automated visual content analysis, the visual importance of the environment becomes evident in relation not only to the issue more broadly but also in relation to particularly located protests. Additionally, the repeated cards/cartoon clusters constituted by memes, posters, and infographics indicates a key form of political communication and deserves further exploration through close reading of these particular images.

7. Conclusion

A cross-platform approach reveals the ways that users leverage the tools specific to each platform in expressing issue alignment, whether through location identifiers, media sharing, hashtags, text, or images.

Twitter

Twitter’s free-form user location field enables political expression where location-identifiers have political significance, as in “official” vs. Indigenous place names. The implication for digital methods research is that user location opens a means of exploring issue alignment and political stance for issues with geographical components. While this study focuses on the locations that are immediately relevant to the pipeline issue (i.e. BC, Alberta, and “Indigenous”), further research would indicate how constituents of various Canadian provinces - as well as those in International locations - approach the issue.

Instagram

An analytical challenge - but also an opportunity - is posed by Instagram’s multiple discursive spaces - images, text, and hashtags - each of which allows substantive expression. The project prototypes a comparative approach to these multiple spaces through analysis of each, including linkages between them where possible (i.e. by hashtag-image and text-image analysis). Taken together, these multiple maps reinforce certain issue patterns; at the same time, individual maps also reveal nuances of the issue through the emergence of new information. For deeper understanding of complex issues such as this, therefore, this project reinforces the value of multiple counter-maps, in line with a critical cartographical approach.

Substantively, through these comparisons, we discovered coloniality exhibited through:

  1. Competing ideals within anti-pipeline sentiment between protection of the land (as pristine), ownership of the land (as a Vancouver resident), and stewardship of the land (as already occupied by Indigenous peoples). These differences reflect epistemological and ontological differences that map onto colonial and Indigenous understandings of the land that are at the root of many policy issues in Canada. At the same time, it appears that over time, these specific clusters increasingly coalesce in the heterogeneous protest space, refusing a restrictive colonial divide. For instance, Slices 3 and 5 of the hashtag-image networks trace a very heterogeneous anti-pipeline space where Indigenous rights, green energy, diversifying the economy, government spending, environmental concerns, etc. intermingle. Together, these findings gesture towards further social research regarding the role of the pipeline controversy as a boundary object, whereby various groups may be brought into closer alignment - whether for strategic purposes or at an epistemological level.

  2. Strong links between Indigenous rights and climate rights, though further research is needed in order to explore the particular nature of these connections. For instance, through the text-image networks, it appears that climate change may serve as a means of critiquing the state as a poor “climate leader”; however, it is not clear how this critique aligns with Indigenous rights. As Indigenous peoples are currently at the forefront of many climate dialogues, both nationally and internationally, this warrants further exploration.

  3. Connections to Standing Rock and UNDRIP, which indicate international solidarity around Indigenous issues. These connections also reveal the use of Standing Rock as a way of framing or narrating the Canadian pipeline issue, drawing parallels between the United States and Canada as settler-colonial states, particularly how these states address environmental, land, resource, and Indigenous issues.

Questions for further research

  • On Instagram, do we see different expressions of coloniality according to analytical method?
  • What might be the significance of these multiple counter-maps to those in the pro- and anti-pipeline camps? How might these multiple counter-maps influence policy-making regarding the pipeline?
  • What might we discover by connecting images via Google API data to original posts? i.e. How will connecting images to text enable more nuanced cluster profiling? Will similar cluster patterns - or new ones - emerge if we begin our analysis with imagery rather than text?
  • How do particular visual clusters change over time (i.e. memes, Justin Trudeau, Indigenous self-representation, etc.)? What do these changes reveal about the issue space?
  • Location data was not available through instagram-scraper. To what extent do toponymic politics emerge on Instagram via location hashtags? How do these compare with politics as expressed on Twitter?
  • How do Twitter and Instagram differ as media through which issues are formulated around colonial topics?
  • How do the different discursive spaces on Instagram relate to users and user types? For instance, do certain groups of users engage primarily in visual discourse, whereas others also use text?
  • How should the multiple discursive spaces of Instagram - image, hashtag, text - be analysed from a digital methods perspective? Together or individually? How to approach this methodologically?

8. References

Borra, E. & Rieder, B. (2014). Programmed method: developing a toolset for capturing and analyzing tweets. Aslib Journal of Information Management, 66(3), 262-278.

Crampton, Jeremy (2010). Mapping: A Critical Introduction to Cartography and GIS. London: Wiley.

Gomez-Pompa, A., & Kaus, A. (1992). Taming the wilderness myth. BioScience, 42(4), 271-279.

Green, D., & Raygorodetsky, G. (2010). Indigenous knowledge of a changing climate. Climatic Change, 100(2), 239-242.

Hawkins, R. & Silver, J. (2017). From selfie to #sealfie: Nature 2.0 and the digital cultural politics of an internationally contested resource. Geoforum 79, 114–123.

Highfield, T. & Leaver, T. (2016). Instagrammatics and digital methods: Studying visual social media, from selfies and GIFs to memes and emoji. Communication Research and Practice, 2(1), 47-62.

Howitt, R. (2001). Rethinking Resource Management: Justice, Sustainability and Indigenous Peoples. Psychology Press.

Joceli, J., Rabello, E., Mintz, A., Sanchez-Querubin, N., Ozkula, S., Sued, G., Elbeyi, E. & Cicali, A. (2017). Visualising hashtag engagement: Imagery of political polarization on Instagram. A Project for Digital Methods Summer School 2017: Get the Picture. Digital Methods for Visual Research. Retrieved from https://wiki.digitalmethods.net/Dmi/InstagramLivenessVisualisingengagement.

Marres, N., & Moats, D. (2015). Mapping controversies with social media: The case for symmetry. Social Media & Society, 1(2), 1-17.

Marres, N., & Weltevrede, E. (2013). Scraping the social? Issues in live social research. Journal of Cultural Economy, 6(3), 313-335.

Rogers, R., Sánchez-Querubín, N., & Kil, A. (2015). Issue mapping for an ageing Europe. Amsterdam University Press.

Rose, G. (2016a). Rethinking the geographies of cultural ‘objects’ through digital technologies. Progress in Human Geography, 40(3), 334-351.

Rose, G. (2016b). Visual Methodologies: An Introduction to Researching with Visual Materials. London: Sage.

Star, S., & Griesemer, J. (1989). Institutional ecology, ‘translations' and boundary objects: Amateurs and professionals in Berkeley's Museum of Vertebrate Zoology, 1907-39. Social Studies of Science, 19 (3), 387–420.

Szeman, I. (2007). System failure: oil, futurity, and the anticipation of disaster. South Atlantic Quarterly, 106(4), 805-823.

Szeman, I. & Boyer, D. (Eds). (2017). Energy Humanities: An Anthology. Baltimore: John Hopkins University Press.

Tsosie, R. (2007). Indigenous people and environmental justice: the impact of climate change. U. Colo. L. Rev., 78, 1625.

Tuck, E., McKenzie, M., & McCoy, K. (2014). Land education: Indigenous, post-colonial, and decolonizing perspectives on place and environmental education research. Environmental Education Research, 20(1), 1-23.

Venturini, T. (2010). Building on faults: How to represent controversies with digital methods. Public Understanding of Science, 21(7), 796-812.

Venturini, T. (2010). Diving in magma: How to explore controversies with actor-network theory. Public Understanding of Science, 19(3), 258-273.

Venturini, T., & Munk, A., & Munier, A. (2018). Data-Sprinting: a Public Approach to Digital Research in Routledge Handbook of Interdisciplinary Methods (Ed, Lury, C.). Routledge International Handbooks.

-- CarrieKarsgaard - 02 Aug 2018
Topic revision: r5 - 09 Aug 2018, CarrieKarsgaard
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback