The Hysteriometer: 8 Years Later

On waking up this morning, a glance at my phone notified me that my "most commented" post on Facebook from 2012 was on The Hysteriometer Plot, a visual from an online-hosted Application developed by a then-colleague, Hugo Carr (https://www.dhirubhai.net/in/hugocarr/)

I can't find the code that underpins the chart. I have found an archived blog post from 2012 inauspiciously entitled "Learn MATLAB Plots Using Trendy" which links to an online Independent article entitled "Calm Down, Dears" which outlines the method behind the plot:-

"Hugo Carr ...... went through the front pages of newspapers counting the total number of words and dividing by the number of emotive words such as "outrage", "scum", "hysteria" etc, to give each publication a hysteria rating."

This was 2012. How the technologies behind this app have transformed our technical and social worlds in the 8 years between then and now. Here are some.

Technology Impacts

Alternative Data: The premise of the Hysteriometer app was linguistic programming, in this case measuring sentiment that tends to hysteria. Sentiment has driven the proliferation of "Alternative Data", a term that started in 2014, so says altdg.com. I can't quite recall when I first heard the term used, but 2014/2015 has been referenced to me by several quant friends as when "alternative data starts". Eagle Alpha, a current lead in AltData formed in 2012, but may not have used "Alternative Data" term at the outset. Qandl launched its first data-sets in 2013. Sentiment data provider, Ravenpack, had been around for far longer working with Dow Jones to assess and quantify news sentiment: I recall that they mobilized around "sentiment analysis". In essence, the emerging days of alternative data was led by the quantitative linguistic categorization of news, earnings announcements, central bank statements and other forms of sentiment analysis. This has underpinned the financial service industry's proliferating alternative data explosion. Simple hysteria for some, financial meaning for others.

Internet-of-Things: The "Trendy" platform alluded to in the blog title was a server that executed code which connected front page web-sites, in this case to a sentiment analysis program which outputs a jpg for a web-page, early stage IoT. Indeed, if you follow through the article links from 2012, where you would have once been passed to "Trendy", you now get forwarded to a much more fully-fledged IoT platform. Now IoT is everywhere, not monitoring newspaper hysteria but controlling household energy consumption, driving analytics from satellites, weather stations and other physical data/image acquisition devices for cloud post-processing and information deployment. The platform which ran The Hysteriometer in 2012 has morphed into a whole technology area that affects all our lives.

"Trendy" is No More. And thank heavens for that, not just its awful name and the gut-wrenching whooping and hollering that I recall from the "kick-off" where it was announced alongside its' evil twin "Cody". Trendy was an example of an early effort by commercial organizations to engage with "citizen data scientists" and ideally promote a service of some sort at the back-end to pay for it and then some. These were also early days for the likes of Tableau and Qlik on the dedicated solution side with no sign of PowerBI or a useful Google Analytics. Of more lasting impact though, R was approaching peak hype in 2012 with RShiny (the online and server-side visualization platform) gathering momentum. To be sure, software communities were coming together to share applications outside corporate sponsorship. However, the most interesting collaborative tool the Jupyter Notebook, for full-on script sharing and execution too, was yet to appear. Indeed, early Python adopters in finance and data science were only just getting started, in Fin Svcs largely limited to the leading/bleeding-edge systematics such as AQR, DE Shaw, Man Investments. While Trendy itself was rather naff and its commercial and open source descendants do what it did so much better, the basic premise of a live connected application launching new hysteria data every day to the Independent and others was pretty cool.

Human Impact:

UK Press: The Hysteriometer then as now shines a light on the appalling weasel tactics of the UK press. It coincidentally popped up in my feed the day after UK Caroline Flack celebrity was found dead. The immense pressures she and countless others face in the public eye from press-created hysteria have very real human consequences. The "objective" free UK press feeds a vicious cycle of gossip, tittle tattle and artificial outrage to feed the coffers and influencer ratings of capitalist magnates and political power-brokers. As fellow celeb Laura Whitmore put it, "To paparazzi and tabloids looking for a cheap sell, to trolls hiding behind a keyboard - enough." Sadly, I suspect Whitmore's words will ring hollow, not least because we continue to buy these wicked rags.

Computational Linguistics in Public Life: Alternative data referenced earlier is used for consumption, largely for the sake of profit, but the reconstruction of content to influence political speeches and slogans, earnings announcements, central bank policy statements and business PR has far wider consequences. In one sense, we see its use to program fake news. In another, it involves lowest common denominator message simplification: MAGA, Take Back Control and Leave Means Leave. It's increasingly informing public messaging too, with central banks very carefully scrutinizing policy statements in order to anticipate their impacts on exchange rates and other key indicators. An announcement that a bank is considering a rate change can be more impactful in some instances than an actual rate change. At a micro-level, it's impacting social selling and social media marketing - for example the bland over-the-wall corporate imperative instruction social media post type I blogged about two weeks back, e.g. "Learn about the difference between Approach X and Approach Y", "Call our success engineers right now to make you successful", etc. Marketing press announcements too are written as much for search engines and the web-scrapers these days as they are to, well, announce something.

Fake News: But worse. Since 2012, media/cultural power has been exploited on an industrial scale by Trump, his powerful backers, the Brexit campaign (both sides, one in particular) and many others in politics, business, marketing and communications. The new divide is between those who understand how to use algorithmic power to create and manipulate fear and doubt, and those in the tranches of society that don't understand how their agency is increasingly bounded by algorithmic power. The measured causality of presented news, media content and other articles-at-scale to socially-induced hysteria that this chart conveys are now widely applied/abused by politicians and businesses alike, easily deployed across global online communities for macro-messages and at a micro-scale for community "social selling". The quantitative application of semantics of the likes of Kellyanne Conway and Dominic Cummings have superceded the unmeasured spin of Alastair Campbell and Peter Mandelson.

Note I am not saying this is a right wing/nationalist strategy - they just happen to be better at it than their left-of-centre rivals, for a host of reasons.

Parting Thoughts: Interesting how time changes the relevance of software assets. At the time, the Hysteriometer was a hobby application run on a gimmicky half-baked platform. However, the technical and social impacts of this very superficial, simple example and its underpinning platform are immense. Unsurprising it was my most commented Facebook post in 2012, though to be fair probably my only commented Facebook post that year, maybe ever.