An already interesting blog post about people and algorithms at ‘socializing finance’ was made more so when I noticed who wrote a comment disagreeing with the post — Mark Cuban, tech billionaire, Dallas Mavericks owner, and blogger himself. I know why I’m reading a blog called ‘socializing finance’ — to learn more about sociological frameworks for finance and accounting, and tech for that matter. But a tech billionaire?
The author of the post (Daniel Beunza) wanted to convey that algorithms are powerful, but so are the people behind them:
In trying to convey the full novelty of algorithms, these are presented as a technology that is entirely autonomous from human traders. And that, Frankenstein-like, they might turn against their creators at any time.
In fact, the reality is much different. These algos are constantly monitored by the traders, who tweak their parameters to fit the market context. My point is, high frequency trading has a key social dimension that is forgotten in the article — and in much academic economics. Overlooking this dimension makes algos look more threatening than they actually are, and gives the mistaken impression that the only skill necessary to master them is familiarity with computer code. Behind every seemingly autonomous algo there is a person. But of course it’s more interesting to think of algos as living bacteria endowed with their own soul.
Mark Cuban replied:
you are wrong. Algorithms can be misprogramed. They can be misconceived. As part of an equation, you are right. But when those algorithms control millions of shares of stock, that is a possible recipe for disaster.
When algorithms have to interact with each other, the owner controls only their own and acts like they are in a vacuum.
a problem with one can lead to unexpected results with the other.
its nice to say that algorithms are people too, but that is not the case
To which Beunza responded:
I would say you are incurring in the symmetrical error that you denounce. The key in understanding HFT is avoiding the extreme views that reduce this activity to either humans or machines. By cautioning that I am reducing them to humans (which I’m not, I’m just reminding people that humans are still there), you actually reduce them to machines.
How then to retain a symmetrical view? To do this, I’d argue we need to look at the combined “assembly” of humans and machines. (This has been the key focus of French sociology of innovation for some time.) And focus in things like: what is the interface used by HFT traders, and how much can they see? How to build slowness into the system at key points in time so that humans can provide judgment? How can shame, embarrassment and norms of good behavior be brought back?