In an increasing data-driven society, storytelling is considered a method, even maybe an anti-dote, for making complex technologies (Giaccardi & Redström, 2020) tangible and more human. Storytelling is the result of human activity, which is difficult for machines to learn to master. This might also explain why storytelling has been increasingly of interest of scholars over the last 10 years in the context of academic fields as human-computer-interaction or human-data-interaction. While a large body of work has mainly focused to expose storytelling techniques for data experts and professionals, traditional storytelling areas as film and video has been focusing on participatory storytelling approaches to let the viewers contribute actively in the story. These approaches have been particularly useful to give underrepresented actors a voice as mainstream media often rely on actors that represent a large demographic range (Manni et al., 2019). Such participatory storytelling approaches, however, also holds much promise for the communication of science data, as it establishes trust in the combat to the spread of misinformation (Meijer, 2009).
Typically, science data is objective and also often quantitative in nature. Yet to add meaning to this data, authors often enrich it with qualitative data such as anecdotes, perceptions and experiences (Liu et al., 2020). We consider the participatory input in the story construction to be mainly quantitative (e.g. with sensor data) yet also allow the addition of qualitative data (e.g. perceptions, imaginations). Such approach to data representation is in line with timely studies on making data meaningful (Liu et al., 2020).
In this report, we will connect existing data storytelling structures with those of participatory media. First, we will define storytelling in its traditional, interactive and data-oriented form. Then, we will elaborate on the role of participation in storytelling. Third, we explored these narrative structures and techniques in 17 lo-fidelity data stories. Finally, we will introduce the ParCos Storyteller, which entails participatory data storytelling techniques.
What is data storytelling?
In this chapter, we will define storytelling and the role of data plays through examples in different media. There is a difference between storytelling as a method in which participants construct stories together as a way to voice their opinion and engage with each other (Manni et al., 2019), versus storytelling as a product that is understandable for users that were not involved in the process. Therefore, as mentioned in D2.1 Bristol Approach to Citizen Science, the ParCos Storyteller is mainly concerned with the fifth step of the Bristol approach, i.e. orchestration, or the presentation of information to other stakeholders to trigger additional input data.
In its essence, storytelling is considered as a linear sequence of occurrences. The most traditional narrative structure consists of three-acts, i.e. setup (act 1), confrontation (act 2) and resolution (act 3) (see Figure 1). Storytellers continuously reinvent this structure by shifting the order, adding or removing narrative elements and storylines. A storyline is a narrative thread that can be experienced by one character in the story or a subset of characters. Often, there are several storylines in one story that all contribute to the main narrative goal. A typical example is the parallel editing or cross-cut in which you see two scenes that are intertwined, such as 1) the crook and his/her plans to derail a train and 2) innocent passengers that enjoy their train ride.
The crook versus innocent passengers is also an example of a conflict. Without conflicts, there is no story. This basic rule is known by all storytellers, and can take many forms. In data stories, conflicts can be identified as the assumptions that exist in the viewers’ mind versus the objective data that might reveal differently, in outliers or in causal relationships, to name a few. Here, characters are thus data sets or data points, and to ‘tie’ the characters emotionally to the viewer, there needs to be a personal link.
A conflict can be the actual story or one of the supporting storylines. In Figure 1, each climax is a conflict. Assume this story is: Central Europe will get warmer in the future. The main conflict lies in the belief of the viewer: whether he or she thinks it is true, and why. A first supporting conflict may then be found in the historical data and how this supports the main narrative goal or not. Then, another supporting conflict may be focused on the outliers and how they increased throughout the centuries. As you may notice, these conflicts are not as straightforward as the crook versus the innocent passengers. Instead, conflicts are implicitly connected to the knowledge of the viewer.
Figure 1. Traditional three-act story structure: 1) setup, 2) confrontation and 3) resolution.
The linear narrative structure is the backbone for stories in different media (e.g., novels, films, presentations) and for different purposes (e.g., entertainment, informing, inspiring). Examples of data stories that follow this structure include infographic films, explainers, news articles with information graphics, etc. Here, the role of the viewer is to consume the story that is presented to her. Data representation is used to support the story. The interpretation of the data is made by the storyteller.
With the rise of interactive technologies since the 1960’s, authors have adapted the existing narrative structures to this digital realm. In interactive storytelling, two main structures exist: 1) interactive narrative structures, and 2) open world.
The first consists of different storylines that are interconnected (see Figure 2) and that allow the viewer to make choices to progress the story. Different forms and shapes of such interactive narrative exist. Yet as our point is to explain how this differs from linear storytelling and this structure is most reminiscent of the three-act structure, we decided to use the branched narrative structure as an example.
This interactive narrative structure is most known as different media – even those that are typically not interactive, such as television – adopted this structure. Indeed, television shows today, such as Dora the Explorer and Bear Grills for children but also the Bandersnatch episode of Black Mirror for adults, brought branched narratives to the TV screen with a simple press on the remote-control button. Yet a branched narrative structure is most known by its application in video games.
Figure 2. Branched narrative structure.
Second, an open world narrative structure is a model that is less tangible for those that are familiar with traditional linear storytelling. Here, the story discovery is literally left to the user with little to no steering of the author towards a particular order in storylines (see Figure 3). This structure is often used in game design that facilitate role playing (RPG), and virtual reality (VR), augmented reality (AR) or mixed reality (XR) experiences.
Figure 3. Open world structure.
What is Interactive Storytelling with Data?
For data stories specifically, the publication of the Narrative Visualization framework (Segel & Heer, 2010) caused an enormous amount of attention for the deployment of narrative techniques to communicate to and engage a large audience with data. The techniques provided by this framework are considered as the basis for our ParCos Storyteller, and therefore we will expose what it entails. It positions storytelling on a spectrum between being author-driven from the one end (i.e. the media professional decides what the viewer sees and when) and reader-driven on the other end (i.e. the user may explore freely). On this spectrum, there are three narrative structures:
1) Martini Glass Structure
Figure 4. Martini glass data story structure.
Figure 4 depicts the narrative structure that resembles the traditional three-act story structure most. It starts with an ‘open’ interface in which the user may interact with settings or data, followed by a linear narrative (that gradually builds up towards the resolution) and ends with an ‘open’ exploration of data points.
2) Interactive Slideshow
Figure 5. Slideshow data story structure.
Figure 5 demonstrates a narrative structure that sequentially offers linear storylines, switched with interaction points that allow the user to decide on the pace. The advantage of this structure is that allows the storyteller to set out the different storylines one by one and as such create a clear overview for complex topics.
3) Drill-Down Story
Figure 6. Drill-down story structure.
Figure 6 shows the most ‘open’ structure. Here, the user may freely decide which storyline he or she explores first. The storyteller’s main task is to bring storylines forward that are coherent when connected without chronological order, which is often not evident.
Why interactivity contributes to data engagement
One of the key reasons why interactive data storytelling is able to engage a wide variety of audiences with data is the ability to find ‘yourself’ in the data. Indeed, finding personal connection points with the data story is an engaging factor that is often missing in linear data stories. In traditional storytelling, the author would create a character (fictional or not) with whom we may identify ourselves, and thus find personal links with. However, for data stories this is not always possible to set up. An example that balances this traditional storytelling with personalization through data is Brooke Leave Home (Concannon et al., 2020), see also D2.3. Brooke is the main character in an online film. As she turns 18, she leaves a care institution in the UK and the viewer follows her on her journey onwards. As a viewer, you gain empathy for Brooke as the video shows the girl in close-ups as well as her home, techniques that are typically used by media professionals. Yet Brooke Leave Home is a project that aims their viewers to engage with data. Here, data is used to personalize the film to the actual location you are situated. What Brooke experiences, in other words, can happen next to your own door.
Why Participation in Storytelling Matters
In storytelling, participation can be interpreted in several manners. Often, participation is understood by professionals as ‘users that generate content for them’ (Varghese et al., 2020). However, true participation in the storytelling process occurs when those users are also enabled to take an active role in the construction of the story and thus have the ability to actively collaborate (Varghese et al., 2020). At the very least, users may contribute data, such asmedia they generated, e.g. videos, photographs, audio and texts. A well-known example in popular media are the photographs of weather phenomena that the weather(wo)man uses in the weather forecast (or news item) to exemplify the current condition (see Figure 7 for an example).
Figure 7. Weatherman Frank Deboosere of VRT discusses the weather of the day of recording (May 20, 2021) with a photo of viewer Caroline De Baets in the background.
Another data contribution by users can take the form of citizen science data, e.g. sensors that citizens installed at their home or office and which results they connect to an online platform where a media professional may base a story on. An example is the project Curieuzeneuzen (‘nosey parkers’) in Flanders, in which 20.000 Flemish citizens participated to measure the local air quality and journalists of newspaper de Standaard regularly publish stories based on this information (See https://curieuzeneuzen.be/).
Combatting misinformation through polyvocality
More contribution by the user may be triggered when the story format allows the user to explore potential storylines and also contribute insights on the envisioned story. Such participatory media examples most often occur in social or health contexts in the documentary genre. Complex problems like mental health benefit from the multi-angle view on the issue that is facilitated through the participatory process (Manni et al., 2019). Here, storytelling is both used as a method and outcome. In community building, the empowering potential of participatory storytelling has been widely recognized (e.g. Lambert, 2006; Bromley, 2010), including the Council of Europe (Lewis, 2008). Also, media professionals, such as those active at public service media, understood it as a tool for democratizing media (Meijer, 2013). BBC, for instance, launched their project Capture Wales in 2003 (Maedows, 2003), and the regional public broadcaster of Utrecht launched a weekly TV show in September 2009, which is still in production today. This project ‘U in de Wijk’ (You in the Neighbourhood) was initially meant to counter the negative impact of journalists that overexposed the problems in that area (https://uindewijk.com/). In its ability to allow different viewpoints on one topic, participatory storytelling is a way to overcome such negative framing (Meijer, 2013). However, journalists are not keen to involve others in the composition of their story, which contributes to the sense of misinterpretation of one’s surroundings and perception of misinformation (Meijer, 2013).
The storytelling that is used in interactive documentaries (iDocs) is related to that of participatory filmmaking as the interactivity is intended as an array of choices offered to users that are enabled to exercise control over the materials presented in the documentary (Manni et al., 2019). Interactivity can serve a number of functions within the iDocs text: finding information (either within or beyond the documentary), learning, furthering the narrative, personalizing the documentary, adding to the documentary content, play or search “playfully” for hotspots within an image interface (Nash, 2012). In iDocs, non-linearity is perceived as an opportunity that “allows audiovisual projects to provide elements to complement and enrich it, providing several added values to the global experience of the audience” (Castells, 2011, p354-3). IDocs thus embrace a higher level of complexity that connects to the polyvocality of participatory filmmaking (Manni et al., 2019).
Media professionals are thus increasingly involving their audience in storytelling. Software tools as Storymaker allow those professionals to process the incoming data that come – in contrast to content produced by experts – in large amounts and in messy structures (Claes et al., 2020). Media professionals act as facilitators and may encourage users to express themselves creatively (Manni et al., 2019). As such, just as the role of the media professional shifted to this facilitating role when participatory approaches entered the profession of interface, service and product design in the beginning of the 21st century. Today, the role of professionals in the media domain is broadened to include facilitation, similar to the role of game designers and interactive storytellers that were already accustomed to think of user input.
In Figure 8, the uncertainty of how the data input of the user may affect the story that was envisioned by the author is depicted as interaction points with dotted lines (which differs from the model of interactive data story structures in the middle of the figure). The figure of the participatory data structure on the right of Figure 8 illustrates how participatory data storytelling links to the ‘open world’ story structure that was discussed earlier.
Figure 8. Linear data story structure (left), interactive data story structure (middle) and participatory data structure (right).
Although participatory storytelling is a timely research topic that is also increasingly adopted in practice, its application in data storytelling is still largely unexplored. In the remainder of this report, we will make the concept of participatory data storytelling tangible through examples and practical guidelines.
Participation in non-narrative data representations
In data visualisation (as it is the main form of representing data), scholars have explored participatory approaches. In interactive data visualisations, participation may take the form of writing text data in a wordle format (Viegas and Wattenberg, 2009) or contributing DIY tracked sensor data on a map (Liu et al., 2020). Also, data visualisation workshops where participants meet face to face to discuss the interpretation of a particular data set (e.g. ParCos workshop at C&T conference in June 2021) can be considered as participatory data representation. Here, data is used as a boundary object to facilitate dialogues about the collective interpretation of that data.
What is Participatory Data Storytelling?
Figure 9 presents the different levels of participation in data storytelling, ranging from no participation (consuming) to low participation (actively interpreting data to find new insights) over medium participation (interpreting data to forward the storyline or contributing data to form a new storyline) to actively participating by contributing stories.
Figure 9. Level of participation in storytelling
Examples of Participatory Data Storytelling
Add data storyline
Examples in this category are mostly derived from research projects, both in academic as governmental contexts. Vital Village, for instance, presents 2 examples on their website. The clearest story is about Food Access and combines scrollytelling (i.e. a website that you can scroll through to discover the story gradually) with a map where participants may upload their own data or insights about data. The latter may feed novel storylines within the larger story (see https://www.vitalvillage.org/data-dashboard/food-access). The narrative structure of this example resembles of the Martini Glass, with the difference that the end now also allows for input and not only exploratory interaction. Another example depicts student transportation equities (see https://www.vitalvillage.org/data-dashboard/customize/student-transportation-equity-map). Here, we are immediately dropped in a map where we, as participants, may add photos and annotations to it and as such add additional storylines. Such ‘photovoice’ is a well-known participatory method to reflect on reality. This story is based on the drill-down structure.
Besides the stories that reside online, also offline forms of participatory data storytelling exist. Often such participation is triggered in workshop settings, where different stakeholders are brought together to reflect on a topic through data and their stories. Such stories may be elicited through existing maps that are augmented by participants with cognitive maps that are drawn with pens to reflect the intensity of perceptions (Liu et al., 2019). Storyboards or scripts, i.e. tools that are typically used in traditional video and film making, are another way to trigger reflection and participation in data storytelling (Wang et al., 2019), while theatre, enactment and play are a more embodied way to experience data and participate in the data story (see D3.1 Guidebook on the use of arts-based methods).
Add data to storyline
We already presented the example of Curieuzeneuzen (see https://curieuzeneuzen.be/) as a cross-over between citizen science and journalistic storytelling, which is similar to the project Tuinlab (see https://mijntuinlab.be/), both in Belgium. Worldwide, Zooniverse is the most well-known example (see https://www.zooniverse.org/) with 2.000.000 citizen scientists contributing data and regular data story updates in various news media. Besides the citizen science stories, the popular website pudding.cool often experiments with eliciting responses of their users to feed a data story, e.g. see : https://pudding.cool/2020/07/song-decay/. In this example, users are asked whether they recognize 90’s songs or not, and with these results they presented a story on the popularity of these songs by users that are younger than the song’s origin date. Thus, users that engaged to deliver input were also curious to learn more about the results in a later stage, similar to the citizen science examples.
However, these examples are most often limited to written press that use the contributed data to base their stories on. As a consequence, there is no direct and dynamic relationship between a change in data and the impact on the story(line). Stories are published when it is newsworthy and depending on the rhythm the press envisioned beforehand (e.g. a weekly item, a monthly spread). The underlying storytelling process (e.g. detecting interesting data) is often black boxed. Examples in which the data stories adapt more dynamically when changes in data are made are less common; the digital platform Coronavirus now is one of the exceptions (see https://www.brig.ht/project/comment-representer-les-data-liees-au-coronavirus-dans-une-creation-digitale-en-3d).
Scholars have studied several workshop tools and methods to facilitate participants to contribute data, ranging from personal informatics to household log (Elsden et al., 2015). While such workshop settings are closely guided by the moderator, artistic examples as data sculptures or installations can take a looser approach in eliciting data from its users. Domestic Data Streamers are famous for their participatory data story installations (see https://domesticstreamers.com/). Here, data participation often takes the form of contributing opinions.
Interpret data to progress storyline
Another way to contribute to citizen science projects – and the stories that evolve from it, is to interpret data that scientists capture. AstroSounds, for example, translates vibrations of space to sound and asks citizens to listen, interpret and discover novel stars (see https://astrosounds.be/). The story, then, is a result of the participation of citizens in interpreting data and being aware for interesting data storylines (e.g., by finding and reporting outliers).
Interpret data to discover storyline
At the very least, the interpretation of data to discover a storyline that was not made explicit by the authors is also considered as participation. In other words, the active discovery of data insights or storylines by the user is a form of participating in the interpretation of data, which contributes to data literacy skills.
Facilitating users to interpret data is not always evident. Data is often connotated to be difficult, too large-scaled, too expert-driven, etc. As a result, scholars have explored ways to present data in more engaging ways, including data comics (Bach et al., 2015), public interventions (Claes et al., 2013) and video (Amini et al., 2015). Recently, scholars studied the impact of data-driven personalised films as a way to familiarise viewers with data concepts (Concannon et al., 2020). Here, users may first consume the story that is tied to their personal living area and afterwards explore how the story changes when moving to other areas.
Consume data story
When data stories are created for consumption, it does not require participation of the user. Yet we consider this level in our model as it is the entry point for users to get acquainted with data stories. Here, examples range from infographic films to supporting charts in a news article. This level also includes science stories that are not literally incorporating data representations in their story yet allow the user to consult their data source.
Reference and source document
Claes, S., Van Den Bosch C., Peeters N. ParCos Storyteller, Deliverable 3.3 of the Horizon 2020 project ParCos, EC grant agreement no 872500, Lappeenranta, Finland.