Below is a list of questions to serve as a starting framework for the discussion in this thread:
- In what instances is data visualization useful and beneficial?
- How does a human rights organization prepare and plan for this?
- Are their formats for data visualization (interactive, etc.) that are better suited for certain situations?
- How can data visualizations be optimized for impact?
- What are essential process elements for the development of data visualizations to be used in human rights advocacy?
- Who should be involved in this process?
- Is there a recommended or ideal process to follow?
- Share stories of success: Implementation of data visualizations in human rights advocacy, and examples of data visualizations that have been optimized for impact
Data visualization is useful in almost all situations where there are abuses of power, data has been collected on these problems and we have access to these datasets. Data visualisation can give audiences access to vast amounts of information in ways that help people understand quantities and relationships. It is a visual method that enables almost instantaneous communication. It can highlight the most critical issues and problems in ways that can inform, educate and also facilitate the creation of more effective strategies for advocacy, policy, etc. I am attaching a screenshot of a variety of data visualisation methods from Google Images.
Rather than looking at the wide variety of instances where data visualization is useful – potentially it will be more helpful to think about where it is not useful, or the limitations of the method. This is a much more interesting question that I will come back to later...
Hello everyone!
I'm glad to be participating in this exchange. I am drawn to begin with a process-related question because, in my experience, the process of creating a data visualization is as valuable to human rights work as as the end product. It requires participants to: 1. consolidate their understanding of their own data (if they conduct original research) and that of others working on the same topic; 2. explore the data to discover a strong advocacy story or message (or test that a preconceived message is supported by the data); 3. delineate clear impact goals and a clear target audience; 4. gain comfort with the creative process, visual thinking, and new tools.
In order to unlock the power of the process, our first job as people familiar with data viz is to encourage potential collaborators to place the right questions at the center. Frequently, we are approached in a manner similar to this: "We are finishing a report on [x topic]. How much would it cost to produce [y infographics] to publish in the report? And can you do it by [z deadline]?"
These logistical questions are natural and practical, but if we want to create powerful data visualizations and data stories for advocacy purposes, it is essential to lead with goals - not outputs or logistics.
Below are some questions I would ask any human rights organization that expresses a desire to enter a visualization process. I invite other conversation leaders to add the questions they use to unlock the most potential at the beginning of the process.
Warmly,
Jessica
____
Can you share your data with us? What other resources are you aware of that may compliment your main sources? (Note: sometimes, partners have a strong story they want to tell, but the data is limited. Other times, they have data, but aren't sure what the message is. Each of these scenarios requires different steps and has a bearing on what kind of output we envision).
What are your goals as an organization, and how does this process fit into that?
Who are the stakeholders? And which of those is your target audience? (Note: it is helpful to define the target audience as narrowly as possible, as it is usually difficult to serve everyone equally).
Where and how do you hope to use the visual materials?
What are the top-level messages that you want people to take away when viewing this work?
Let's say the target audience receives the mesage. What next? Is there an action we are trying to promote? (this question helps us hone in on impact targets)
What do you feel would surprise people about this issue/data?
What factors may constrain this process? (funding. time. human resources, etc) - this is where it is okay to return to logistical questions.
One thing I have learned is that it’s important to think about tailoring your use of data visualization to different target audiences. Research suggests that people with polarized attitudes toward a human rights message (either very positive or very negative) can react differently to the same visualization. In an experiment carried out by an NYU research team (including John Emerson--another discussion leader and me, among others), we found that:
These findings could be used by advocates to ensure that audiences that are likely to be hostile or less open to a human rights message—such as lawmakers or other “opponents”—are provided with tables that they can explore. More friendly audiences might be best reached with more creative data visualizations.
Read the whole paper here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2474695
Check out our resources here: http://visualizingrights.org/
Meg - thanks for this paper. I think there is more research needed on how visualizations impact, persuade and emotionally engage viewers/readers so I'm very happy to add this to my reading list.
I was just teaching a Data Visualization class today (Journalism program) where we went to Fathom Information Design on a field trip. They talked about the complexities of designing data visualizations for different audiences as well as different viewing contexts (phones, printed posters, large touchscreens, web browsers, etc). In one case they had a client who wanted their advocacy project tailored to millennials and policymakers which was extremely challenging for the designers since the needs of those two audiences as well as their viewing contexts are almost polar opposites.
One thing I wonder about the persuasiveness about data visualizations is around what I've characterized previously as their "rhetorical power". Basically, visualizations speak with the language of authority and (at least seem to claim that) they show the "whole" picture. Perhaps their perceived neutrality and authority, as well as the fact that they are perceived as needing a high degree of expertise to make lends them their rhetorical and persuasive power.
Best,
Catherine
Hi everyone,
Great to be able to participate with all of you, old friends and new acquaintances alike. I will probably keep harping on this issue throughout the week, but in my eyes, one of the major concerns for any human rights organization contemplating data analysis and/or visualization is the challenge to resist publishing analysis based on bad data or bad analysis with good data.
With any analysis, everything begins with the data collection methodology. What you can say - and how you say it through visualizations - is inextricably tied to the original data collection methods. Unfortunately, we work on a subject, human rights violations, that is often hidden. We are often challenged to quantify something that is not easily counted. The job of most human rights organizations is not to be the official counter, surveyor or database. And governments still have a long way to go with their open data efforts - especially concerning abuses they themselves commit. This leaves us in a situation where there is minimal data available but a huge appetite for data.
It can be tempting to sate that appetite with whatever data we can get our hands on. This is often data gathered by activists through a wide variety of methodologies. Perhaps it is a sheet of known violations to the organization organized by date. If this data is visualized, there are two, of many, concerns that the organization must think through. A.) Is the analysis and the presentation of the analysis methodologically sound given the data collection methods? B.) Even if it is presented with a long list of caveats, will it be interpreted correctly and what is at risk if it is released and generalized or misconstrued?
The legitimacy of any human rights organization is tied to their devotion to getting the facts right and to not misrepresenting the truth. If a single fact can be challenged, the entire organization can be challenged. And quantitative data is easily challenged on methodological grounds, definitional grounds, and analytical grounds. It is important for us to encourage others to never play fast and loose with data. We need to counter the idea that data visualization is always useful. It can be devastating if it is inappropriate given the original data collection.
Brian, your post warning about the importance of getting the facts right is extremely important. This is perhaps the most crucial element that I try to tell my journalism students about who are just starting out with data. The way I say is usually "Don't trust the data" because they do not typically collect the data themselves. I stress context, collection methods and the interests and incentives of people doing the collecting (why do they collect that data? how is it used internally? what decisions do institutions with that data? whose lives do those decisions impact? what data are they not collecting? why?)
I'm also reminded of a stellar presentation by Patrick Ball who does a lot of work in the human rights space where he walks the audience through the complexities of measuring how many civilian deaths there are in Syria. For anyone interested, this is totally worth it to watch:
Digital Echoes: Understanding Patterns of Mass Violence with Data and Statistics
https://datasociety.net/events/databite-no-74-patrick-ball/
Happy to join this conversation with you all,
Catherine
Yes, our good friend Patrick is the leading voice on this.
"Don't trust the data" is a good framing.
I don't want to hijack our data viz conversation and turn it into a responsible data analysis conversation - but the two are completely linked. A good resource for responsible data analysis, from analytical, security and privacy standpoints is the https://responsibledata.io/ site. The partners putting the site together have done a great job compiling resources.
Some wonderful points in this thread! These questions are only going to get more prominent as data visualization tools become more mainstream and the barrier to data visualization lowers. There are good practices we can follow as data visualizers to provide greater transparency about the decisions we make in analyzing the data, the questions/gaps/uncertainties we faced, our sources, our design choices, and so on. Of course, this transparency is not always possible when it comes to sensitive data. These are some of the nuances that emerged at the Responsible Data event on data visualization, and I second Brian's recommendation to read further on the RDF group page he linked.
From a developer and designer's point of view I definitely agree. It is vital that the communication between the HRO/NGO and the design/development team is working well, and that both sides understand the potential pitfalls. This is especially true if the design/development team is an external team - if so, they should be well informed of the data (and the story) really understand it, prior to sketching designs.
At HRW, the quantitative analyses we use are almost always in the realm of descriptive statistics. Sometimes we will look at associations. We are typically not doing causal or inferential stats. With this in mind, most of the time, quantitative data is used to show one of three things:
I'm attaching two examples from our recent report calling for the decriminalization of the personal use of all drugs in the US: https://www.hrw.org/report/2016/10/12/every-25-seconds/human-toll-crimin...
These are just two graphs, of many, from the report but taken together, they present one of the main arguments: that people of all races use drugs at similar rates but are arrested for use at very different rates by almost every state-level jurisdiction.
The first graph show rates of drug use in the past year broken down by race and drug type. We used past year, rather than lifetime use, to get at current users. The viewer should take away that for all drugs, other than marijuana, black, white and latino residents use drugs at the same rate. (As an aside, this would be a good graph to critique for showing uncertainty. We used box plots to show the range of where the true rate lies but I find box plots very difficult for lay readers to grasp. Maybe John could suggest alternatives?).
The second example which combines all three bullets I noted above. It is about rates of arrest for drug use/possession. It compares US states (geographic trend) on the disparity between the black and white arrest rates (comparisons between groups). It is intended to provide an idea of the overall scope of the race disparities, as well as geographic disparities, in how police enforce laws criminalizing drug use.
People read best what they read most, and visual literacy is a hard one. For an article on citizens using design for civic action, I interviewed one of the original designers behind the ubiquitous Nutrition Facts label. The team tested over 30 variations before ultimately abandoning charts altogether. They found that literacy and visual literacy are more complex than they had imagined.
One of the great powers of statistics is the ability to work with uncertainty, but presenting this visually is a particularly thorny problem. There are some references and papers here on visualizing uncertainty and its challenges.
One thing that might soften the box plots in these specific charts might be to switch from a hard outline, which over emphasizes outer bounds, to a solid shape. Or you could go even further and blur the top and bottom edges with a gradient. You might also consider trying a violin plot if you have the appropriate data.
I find the varying y-axis from chart to chart and non-zero baseline here a bit frustrating, as well. I might try to use the same scale from 0 to 20 for the first three, perhaps 0 to 2.5 for the latter three and see what that looks like. Your key take-away, that the rates of drug use are not very different, might even be emphasized.
It may be worth touching on the difference between data visualization and other popular and powerful visual tools at this point (such as infographics). Brian's visuals above are solid examples of data visualization, as they:
For advocacy purposes, I would argue that there is value in framing our findings within a much more direct storyline or message in some situations. Brian's visuals above, for example, require a few minutes to absorb, the willingness to read accompanying analysis, and decent chart literacy. We cannot expect every audience to engage on these terms, and this can be limiting for advocacy purposes. Perhaps an interesting question for group discussion would be: to what degree do you agree with the statement "all good data visualization must be objective"? Can anyone share examples of powerful/impactful data visualizations that take a stronger hand in guiding viewers toward a specific conclusion?
I would also like to point out that data visualization is a term often applied loosely by newcomers to the field. For example, infographics, data journalism, and data visualization are terms frequently used in the same breath/interchangeably. Infographics are very popular and common advocacy communication tools, but they differ from data visualizations in a number of key ways:
In an advocacy context, I have worked worked on both infographics and data visualizations, but more of the former. They are different tools and they fit different goals. I won't go into that too much further, since this is a conversation about data visualization specifically, but I hope it is helpful to consider what data visualization is not, as well as what it is.
Personally, I don't agree with the statement "all good data visualization must be objective". I believe a good visualization should tell a story, which in many cases are subjective. As I mention in another thread (https://www.newtactics.org/comment/9338#comment-9338) I differ between exploratoty and explanatory visualizations, and provide examples of what could be considered objective and subjective visualizations. Exploratory visualizations could provide interaction functionality to explore the data in an objective sense, though they should still be centered around the story (at least in an initial state). Explanatory visualizations could very well be subjective.
It's worth pointing out that being subjective still assumes you stay true to the data and don't mislead the reader. You should clarify if the presented data is a "special case", or based on any crucial assumptions, etc.
Thank you everyone for your comments. There are many excellent points and things to talk about on this thread but I want to focus on the limitations of data visualization. I think this is subject worth discussing because, as Catherine says, data visualization can be very rhetorically powerful – and yet it does not always warrant the authority that it presents. Her emphasis on the problems with the perceived neutrality of data visualization is foundational. It seems to me that it is very important right now, with the proliferation of fake news, to help people think critically about what datavis can do well and what it can’t do. Several people have already made comments that fit this theme so I will start by making an overview of their contributions.
Brian’s discussion on the challenge of ‘bad data’ goes some way towards opening up discussions on the limitations of data visualization. He describes the dangers of hidden data; of things that cannot be easily quantified; and the problems accessing data on politicized topics. I can’t emphasis enough the risks associated with the problems he highlights, especially associated with using “whatever data we can get our hands on”.
Jessica has just emphasized the relevance of considering what data visualization is not. She has asked us to consider: “to what degree do you agree with the statement ‘all good data visualization must be objective’". I’d like to explore this idea over the next few days. She also considers the relationship between data visualization and other visual tools (infographics) and the value of narrative for advocacy (best developed with an infographic approach). I agree with Jessica’s emphasis on narrative. I think it follows from an understanding of the limitations of data visualization.
I find John’s categorization of the three areas/uses of data visualization a helpful starting point for thinking about strategy. He describes data visualization for: 1) analysis and exploration. 2) communication and advocacy and 3) facilitation and collaboration. If we were to discuss this categorization in more depth I think we might find that other types of visual methods will be necessary for all three areas (following an examination of what data visualization cannot do).
Finally, Meg I am interested in reading your paper and I hope I will be able to do so before end of the week and comment. This is a subject I want to think about a little before I delve in.
I have just run out of time this morning but if you have time and an appetite for more on this subject I have recently written a conference paper called Data Visualisation Does Political Things. I will be back later...
Joanna,
Thank you so much for sharing your conference paper, Data Visualisation Does Political Things. This very powerful statement is central to human rights advocacy (found on pg 6):
"It is worth repeating: data is not neutral. It is instead an assemblage of infrastructures, laws, social discourses, technologies and politics (Corby 2015). (emphasis mine)
Your paper helps to point out that data visualization is a powerful tool, but with limitations and caveats that must be addressed by human rights advocates.
Your comment regarding, "decisions to use data (in a plethroa of different ways) serve political priorities and agendas that must be recognised." As human rights advocates, we have specific objectives that we want to advance and our ability to understand what data is collected (by who, through what mandates / laws / institutions / organiations, etc), or data NOT collected (the decription you use for darkdata), can provide human rights organizations with a better grasp of what data is missing, evaluating if they have bad data (see Brian's post), or how to interpret the way others are using data.
I was particularly intrigued by this breakdown of understanding that you presented in the paper (pg 7) :
Thanks for your comments Nancy. I agree that Jessica's post on shifting the question as a good place to start. Thinking critically about data visualisation on politicised topics involves considering the narratives that are obscured due to power imbalances in society. We can start with the discourses, narratives and stories that are already being told about a situation and compare these to the narratives we want to tell. We can try to figure out why these are different. Whose interests are being perserved in the dominant story? Whose interests are denied? Then we are in a better position to look at what stories are not being told and why they are not told.
I also want to flag up some recent articles that raise concerns that statistics themselves will become less reliable under the new adminstration in the United States. It is hard to make data visualisations on human rights abuse if there is no data to support it. These two articles point out alarming trends: