The Newest Viral Media Graphic
Updated: Dec 22, 2019
In November 2017, I stumbled upon Vanessa Otero’s viral media graphic, which, at the time, was in its third iteration. In my first blogpost on this subject, I explained why I found the graphic intuitive and appealing; I distilled her more detailed set of seventeen of evaluative criteria into my own broader set of five, sometimes overlapping, categories; and I offered a theoretical explanation of the curve’s apparent Gaussian shape, resembling a standard probability distribution function. I further offered some suggestions related to issues of dimensionality and replicability.
Since that time, Ms. Otero has made tremendous strides in further developing her graphic, including creating a new blog, ad fontes media, Latin for “to the source.” The newest version of the graphic, Version 4.0 (reproduced above), might appear slightly less appealing visually, mostly in the sense of being cluttered, even as its essential features remain unaffected. In fact, however, the clutter is part of Ms. Otero’s ongoing plan to render the graphic more comprehensive, with larger numbers of included media sources, greater moves toward replicability (more below), and a planned vehicle for individual media source searchability. This is all extremely ambitious, and I think a great service to her many followers.
Here I will offer some reflections on the earlier themes, with added insight from the newer and more developed graphic. I will discuss the following points: (1) the dimensionality of the graphic; (2) the shape of the graphic; and (3) replicability of data within the graphic. The first two points might sound similar, but they remain different in important ways.
The dimensionality of the graphic:
In her detailed post explaining this newer version of the graphic, Ms. Otero tackles an important question raised by some of her blog readers. Is it sound to assume a conventional left-right—extreme liberal to extreme conservative—ideological dimension or, instead, do some media call that assumption into question? I begin here because the question relates to a major theme of this blog, where I have paid a great deal of attention to the dimensionality of our larger politics, and because, despite notable differences, the divisions among our media in some respects is a microcosm of the divisions within our polity. Libertarians, for example, have questioned the single ideological dimension represented in Ms. Otero’s media graphic. At least nominally, libertarians eschew government intervention into economic liberties (skewing right) and into individual privacy, including especially decision making with respect to matters of sexual autonomy (skewing left). In addition, a notable segment of the electorate, even if a minority, appears to pull both ideological extremes inward, like a horseshoe, possibly rendering an inside-outside dimension more descriptive of the relevant spectrum than the conventional left-right presentation suggests, at least with respect to some important issues. Ms. Otero’s media graphic, however, retains the conventional ideological spectrum presented along the horizontal axis.
In an earlier post, The Dimensionality of Trumpism, I posited that both the Bernie Sanders and Donald Trump campaigns presented a potential insight-outside challenge. I further observed that after Trump took control of the Republican party, the dust would eventually settle, and each of the two major political coalitions, associated with each of the two major parties, would coalesce around revised endpoints. In three-dimensional space, based on Trump’s having prevailed both in securing the Republican nomination and in winning the general election, the effect could be to make the Republican party either more or less conservative, or more or less prone to eschewing the traditional right-left evaluative criteria. (The same might have been true at the other end of the spectrum had Sanders been successful). I also explained, the manner by which we choose our head of state ensures that in the long term, a two-party system will remain a dominant feature of our politics. This implies that at whatever locational point the Republican party settles, that will form one end of a revised spectrum, depicted linearly. The other endpoint will represent the also-shifting Democratic party coalition. Within a three-dimensional space, the two points form a line, and wherever that line is, and in whichever direction it points as compared with the line it supersedes, becomes the basis for a new left-right coalition.
To this observer, there’s little doubt that we are experiencing changes, perhaps notable ones, in each major party’s dimensional location. During the resulting period of disequilibrium, it is hard to discern precisely where each party will settle within the three-dimensional space. However this shakes out, a single-dimensional ideological spectrum will emerge, albeit one that incorporates elements that challenge the locational premises of the single-dimensional spectrum it supplants, and that has characterized much of our politics from roughly the end of World War II through the Obama Administration.
For purposes of the viral media graphic, these observations might appear to present notable challenges. Consider two hypotheticals. When a traditionally conservative commentator criticizes Trump’s treatment of foreign leaders, policies on international trade, and personal conduct, and is even willing to endorse a Democratic candidate with the goal of ensuring political and financial stability at home and on the global stage, is she operating on the left or right of the spectrum, or, instead, is she eschewing the outsider-insider dimension, hoping to restore the left-right spectrum? How about a traditionally liberal commentator who eschews the rise of the new socialist wing of the Democratic party, perhaps on ideological grounds, or perhaps for fear that it will ultimately badly splinter that party, thereby inceasingly the likelihood of a Trump victory in 2020? The answers are not obvious. It is possible that the conservative and liberal have flipped sides, yet it seems more likely that each has chosen to eschew the outside-inside dimension today, hoping to revert to her or his original ideological positions once the traditional left-right spectrum, or something quite like it, is restored.
The uncertainty respecting the ideological locus of some substantive policy positions expressed in the media (or otherwise) should not be viewed as an insurmountable obstacle to Ms. Otero’s project. It might be impossible, or at least impractical, to code across multiple dimensions. Even if Ms. Otero could do so, there are reasons to suspect that over time, our politics will eventually settle along a single dimension, albeit one that might look different in some ways from that which has historically dominated. At worst, she might mistake a position as left or right that, in a later period, will reposition to the other side. "The perfect is the enemy of the good." And Ms. Otero's project is better than good. More to the point, her project is more about the process of coding than the final product of an immediate graphic. This is, after all, version 4.0, and I suspect, and hope, that there will be many more to come.
One more observation about dimensionality: This newer graphic sharpens the criteria along each axis. Whereas the prior version contained qualifiers on the left right axis that appeared qualitative, rather than purely ideological, this graphic cleanses each axis of potentially conflated criteria. It is clear from her choice of characterizations, which appear on the graphic itself, that the left-right axis is intended to capture liberal-to-conservative ideology, and that the up-down axis is intended to capture credibility to non-credibility.
The shape of the graphic:
Some commentators questioned my observations about the prior version, 3.0, and its apparent Gaussian shape. A probability distribution function plots data on a graphic to generate a distribution. Depending upon what is being plotted, for example when grading a group of exams, it is often intuitive to imagine convergence toward the mean, with a descending data population along either side, until the tails, left and right, are sparsely populated.
Ms. Otero’s graphic is not generated with data plotted in this manner. The media at the top center are not there because there are many of them. Quite the contrary. The familiar shape of the curve thus demands a theory separate from population density hovering at the mean, and although I offered some suggestions in my last post, I did not sufficiently clarify that relationship. A clearer account rests on the intuition of seeking to maximize the relevant output—measured in terms of readership, advertising revenue, profit, or some other relevant metric—along either dimension for a fixed location along the alternative dimension.
The graphic depicts a series of apparent equilibria points for successful media across the dimensions of quality (or accuracy) and ideology. A media outlet maximizing, for example, profit or readership, will optimize across these two dimensions. As a media source approaches the mean ideology, neither skewing left nor right, this implies that there is little ideological bias, rendering the source more credit worthy. As a media outlet departs from mean ideology, it is apt to descend along the dimension of creditworthiness. This also holds true in reverse. As a media outlet is increasingly credit worthy, it is less apt to publicize unsubstantiated stories that compromise that reputation, and, conversely, a less credit worthy media outlet is more apt to publicize stories that draw in readers tending to favor its ideological perspective, however controversial and unsupported those stories might be.
The graphic reveals a broader spectrum of tradeoffs comprising the resulting equilibrium points. In between the extremes, media make relevant tradeoffs that reflect the descending left and right slopes. For example, some media that emphasize commentary but that lack the resources for original fact investigation, will exhibit some bias, but will nonetheless avoid unnecessarily salacious stories or other stories that are unsubstantiated by credible sources. Other media are more willing to sacrifice the commitment to verifiable sourcing as the means of attracting readers committed to a shared ideology. And, of course, there are still finer gradations.
In the prior post, I stated that Ms. Otero’s evaluative criteria largely distill to a list of factors associated with credible bonding commitments signaling a higher probability of accuracy, on one dimension, and ideological commitment, on the other. As previously noted, her newer evaluative criteria are further consistent with this interpretation, and it is the relationship between these criteria that gives the graphic its shape. Consider this: for any given ideological commitment (equidistant left or right of the median), what is the maximal accuracy a given media source can afford to ensure? Conversely, for a given level of resources to deploy at credibility bonding, what is the maximal departure from a neutral ideological positioning that it can achieve? The shape of the graphic represents a meaningful set of equilibrium points based on these tradeoffs for the most media outlets.
Of course, there are some notable exceptions, including CNN, a credible source that in Ms. Otero’s graphic is further south than one might assume, due largely to news repetition, and the National Enquirer, which although relatively neutral on ideology overall, suffers notably along the dimension of credibility. Economists would describe these seeming off-the-curve media as providing differentiated products. CNN is not just about credibility; it is about ubiquity, a constant, yet credible, 24/7 source of information. As such, CNN is willing to couple redundancy with carefully researched or sourced delivery. Witness the never-ending alerts to “breaking news.” Similarly, setting aside the close Trump linkages, the National Enquirer pulls in readers with unrelated, and often outrageous, stories, the stuff of supermarket tabloids, while also occasionally catching a scoop, or the reverse, suppressing one via catch and kill.
Finally, in this graphic, Ms. Otero adjusted the earlier unequal distancing along both axes. By making the spacing equidistant, it has somewhat compressed the earlier visual effect, giving it a less pronounced Gaussian shape as compared with version 3.0. In addition, along the relevant dimensions, left and right, there are media outlets occupying each of the identified positions starting with Original Fact Reporting, at the top, and descending to Contains Inaccurate/Fabricated Information, at the bottom. In this newer version, along the right side of the graphic, Ms. Otero offers helpful general categories that correspond to newly imposed colored rectangles, each capturing general media cohorts based upon the combined dimensions.
Replicability of data within the graphic:
Ms. Otero's new website is also the start of a newer method of generating data. However superhuman Ms. Otero might appear to be, she’s only one person, and, along with the rest of us, a person with admitted biases. Her goal is to find left-center, right-center, and moderate readers to evaluate the relevant media stories, and to see how the process of averaging the reviews fares, for example, as compared with her own analysis. Ms. Otero offers up two ways to evaluate for the left-right axis, one substantive, one linguistic. Her eventual goal is to see whether it’s possible to rely upon artificial intelligence to perform this coding function. If so, this would allow her to test overall replicability without the massive resource investment associated with human coders. (Of course, the AI will require its own coding and hardware investments).
I suspect, although I do not know, that this task of AI coding will prove more manageable for linguistic than ideological inputs. Key words are more searchable than substantive ideas, which can be subtle and evasive, and which can be expressed in either ideological direction using similar vocabulary. Even so, this is a highly worthwhile experiment, and time will tell whether it is manageable.
Ms. Otero has made clear that her goal is to evaluate the source (hence “to the source”) not individual writers. And at least at this stage of the project, that’s a sound judgment. She is also now engaging in the separate assessments of individual shows within such media as CNN, MSNBC, and others, rather than limiting her ranking to the networks. This will help with replicability by breaking down what is being assessed even as it creates more data points to to analyze.
One open question is whether the generally Gaussian shape that remains even in Version 4.0 masks something resembling a more bimodal distribution among media. This is separate from the question of dimensionality. With an actual probability distribution function, bimodalism can arise along a single-dimension using left-right scale. I suspect, although I do not know, that whereas the electorate has proven increasingly bimodal over the past several decades, future versions of the media graphing are unlikely to create the visual appearance of bimodalism, at least to the same extent.
The sharp divisions among the media fit comfortably, as this version of the graphic shows, with a largely single-peaked curve. Again, as applied to the media, this is not a mode as used to describe a probability distribution function. Instead, the question is whether there is a greater distance than first appears between the top center equilibrium points for the most credible media operating slightly left or right of the median. Notably, Ms. Otero has observed that with respect to the media, it proves easier to identify sourcing to the left or to the right of center than to identify the actual center. In addition, among the central insights emerging from the graphic is that as media outlets increase credibility, they move toward ideological neutrality. These factors also contribute to the graphics shape.
Once more, I consider this project extremely valuable, and I look forward, along with so many others, to seeing where Ms. Otero takes the project from here. A central insight of equilibrium analysis is that the process is dynamic, not fixed. The same holds for the ambitious project of ranking news media.
I welcome your comments.
[Special thanks to Vanessa Otero for comments on an earlier draft and for permission to reproduce Version 4.0.]