Category Archives: Domestic politics and elections

Trustworthiness of unfree code

Several reports are circulating (e.g., via /.) of a court case in New Jersey in which the defendant won the right to audit proprietary genetic testing software for errors or potential sources of bias. It being a murder trial, this is about as close to a life-or-death use-case as possible.

Given the stakes, it is understandable that a low-trust standard should prevail in  forensic matters, rendering an audit indispensable (nor is the firm’s “complexity defence” anything short of untenable). What is surprising, rather, is how long it took to obtain this type of judicial precedent. The authoritativeness deficit of algorithms is a topic of burning intensity generally; that in such a failure-critical area a business model based on proprietary secrecy has managed to survive is truly remarkable. It is safe to say that this challenge will hardly be the last. Ultimately, freely auditable software would seem to be the superior systemic answer for this type of applications.

Turning the page on disinformation?

With the inauguration of a new Administration, speculation is rife on the chances of moving on from the more toxic aspects of the political media ecosystem of the past half decade. An op-ed by Rob Faris and Joan Donovan of the Shorenstein Center (Harvard Kennedy School) spells out these aspirations concretely: with Biden in the White House, conservative media such as Fox News have the opportunity to distance themselves decisively from the more fringe disinformation beliefs of the conservative base, and return political discussion to a debate of ideas rather than the reinforcement of antagonistic social realities. In their own words,

The only way out of this hole is to rediscover a collective understanding of reality and to reinstall the mechanisms of accountability in media where they are missing, to ensure that accuracy and objectivity are rewarded and disinformation is not given the space to metastasize.

I think there is good reason not to be particularly sanguine about these goals. Faris and Donovan’s proposed solutions read more as a restatement of the intractability of the problem. For one thing, their discussion is very top-down, focusing on what the upper echelons of the Republican Party, the conservative-leaning media, and their financial backers can do. The trouble with US political disinformation, I would argue, is that at this point in the cycle it is largely demand-driven: there is a strong appetite for it in the (GOP) electorate at large, to the point that one could speak of a general social antagonism in search of arguments. Hence, focusing on the infrastructure of production of disinformation is merely going to elicit creative responses, such as the flight to alternative social media platforms, which will be viable given the size, means, capabilities, and diversity of the public involved.

The alternative, however, is equally fraught. Focusing on the transformation of mass beliefs in order to discourage the demand for disinformation amounts, in essence, to a domestic ‘hearts and minds’ mission. The historical record for such attempts is hardly promising. The trouble, of course, is that political adversaries cannot at the same time be treated as respectable dissenters in the common task of running the commonwealth and also as fundamentally wrong in their factual beliefs: respecting and correcting struggle to coexist in the same interpersonal relationship.

One of the problems with such an approach is that it is incomplete to say that the US media ecosystem is fragmented and siloed:

Since its inception, conservative media in America has operated under different rules […] The outcome: a cleavage in the U.S. public sphere and a schism in the marketplace of ideas. The news media of the center and left, with all its flaws and faults, operates in a milieu in which fact checkers have influence and the standards and practices of objectivity and accuracy still hold sway.

In other words, conservatives have largely seceded from the traditional, 20th-century unified media sphere of print and broadcast outlets, toward a smaller, insular, homogeneous but culturally dominated one of their own. The rump ‘mainstream media’ has maintained its old ‘fourth estate’ ethos, but not its bipartisan audience. Hence, its loss of cross-party authoritativeness.

The accountability void created by this partisan segregation of US public opinion offers concrete inducements to ambitious populist politicians, which will prove hard to resist. The belief that the system contains self-correcting mechanisms appears ill-founded. Yet, it is unclear that the current administration has the stomach for the protracted effort necessary to change mass beliefs, or that it would be supported consistently by external power centers, especially in the business community, in doing so.

Victory lap for the EIP

Today, I followed the webcast featuring the initial wrap-up of the Electoral Integrity Partnership (I have discussed their work before). All the principal institutions composing the partnership (Stanford, University of Washington, Graphika, and the Atlantic Council) were represented on the panel. It was, in many respects, a victory lap, given the consensus view that foreign disinformation played a marginal role in the election, compared to 2016, also due to proactive engagement on the part of the large internet companies facilitated by projects such as the EIP.

In describing the 2020 disinformation ecosystem, Alex Stamos (Stanford Internet Observatory) characterized it as mostly home-grown, non-covert, and reliant on major influencers, which in turn forced platforms to transform their content moderation activities into a sort of editorial policy (I have remarked on this trend before). Also, for all the focus on social media, TV was seen to play a very important role, especially for the purpose of building background narratives in the long run, day to day.

Camille François (Graphika) remarked on the importance of alternative platforms, and indeed the trend has been for an expansion of political discourse to all manner of fora previously insulated from it (on this, more soon).

Of the disinformation memes that made it into the mainstream conversation (Stamos mentioned the example of Sharpie-gate), certain characteristics stand out: they tended to appeal to rival technical authority, so as to project an expert-vs-expert dissonance; they were made sticky by official endorsement, which turned debunking into a partisan enterprise. However, their very predictability rendered the task of limiting their spread easier for the platforms. Kate Starbird (UW) summed it up: if the story in 2016 was coordinated inauthentic foreign action, in 2020 it was authentic, loosely-coordinated domestic action, and the major institutional players (at least on the tech infrastructure side) were ready to handle it.

It makes sense for the EIP to celebrate how the disinformation environment was stymied in 2020 (as Graham Brookie of the Atlantic Council put it, it was a win for defence in the ongoing arms race on foreign interference), but it is not hard to see how such an assessment masks a bigger crisis, which would have been evident had there been a different victor. Overall trust in the informational ecosystem has been dealt a further, massive blow by the election, and hyper-polarized post-truth politics are nowhere near over. Indeed, attempts currently underway to fashion a Dolchstoßlegende are liable to have very significant systemic effects going forward. The very narrow focus on disinformation the EIP pursued may have paid off for now, but it is the larger picture of entrenched public distrust that guarantees that these problems will persist into the future.

Robocalling campaigns

Politico‘s Morning Tech reports that the Trump campaign has launched a poll on its website to gauge sentiment as to Twitter’s anti-conservative bias. There is nothing particularly scientific or informative about the poll. In fact, MT speculates that the main purpose of the stunt is to get respondents to agree, in passing, to have their phone numbers robocalled by the campaign (this kind of data-collection-and-authorization has been done before). Robocalling is one of those annoying-but-effective psychological prods, like canned laughter. Participants in the Twitter poll can safely be considered strong fans of the President, but even they might not consent to being robocalled if asked directly, hence this circuitous route. It is remarkable, though, how outrage is commodified as data harvesting, or –seen the other way– how subjection to invasive marketing is the price of interaction with curated forms of political venting.

Media manipulation

Earlier, I attended an online workshop organized by the Harvard Kennedy School’s Technology and Social Change Project on media manipulation in the context of the 2020 US presidential campaign. Very productive conversation on the tailoring of disinformation memes to the various minority communities. I also learned about the Taiwanese “humor over rumor” strategy…