Tuesday, September 9, 2008

The Trouble Started with a Google Search

From this and similar experiences, Garner realised that people with dementia frequently go back to old memories in order to make sense of otherwise incomprehensible situations.(1)

In today’s news, United Airlines shares suffered heavy losses after business portal Bloomberg passed its 2002 bankruptcy filings as 2008 news. According to the AP report picked up by the New Zealand Herald, the stock lost over a third of its value within five minutes of the story being posted on the Bloomberg terminal and fell as low as $3 (a 75% collapse in value) before the Nasdaq halted trading. Even after the story was clarified and trading resumed, the airline shares still closed over 11 per cent lower than the day before, with a loss of value in the vicinity of 175 million dollars US.

And yes, it all started with a Google search. As the AP explains
A staffer at investment newsletter Income Securities Advisors Inc. in Miami Lakes, Florida, entered the words "bankruptcy" and "2008" in the internet search engine, according to President Richard Lehmann. ISA does this all the time, searching for overlooked information about companies in trouble, he said.

Lehmann said the top item returned in the Google search was a story about United's bankruptcy filing that appeared on the South Florida Sun Sentinel website. The story referred to United's filing on Monday morning. United had filed for Chapter 11 protection on Monday, Dec. 9, 2002. Lehmann said that date was not on the story.
To cut a short story even shorter, the staffer posted a summary on the Bloomberg information service, without so much as checking another source or contacting the airline, and the rest is so much frantic history on the trading floor.

What does this story teach us? We’ve known for quite some time about the power of rumours to lay temporary waste to the financial markets. And lazy journalists (or, in this case, researchers) are not a fantastically new breed either. What seems designed to give us more pause is that somewhat arresting line in the AP story: The trouble started with a Google search.

Of course Google - the company, its technology - is no more responsible for this than automaker X will be when the next idiot full of booze drives off the road and into a tree in a car made by that company. Or is it? Without cars, booze or trees we’d have no accidents involving a combination of the three. And since a tree is unlikely to attack you, and alcohol by itself cannot propel you with the necessary speed to cause the damage, the car might just be the most dangerous ingredient of the lot. Of course one cannot ignore that a person was behind the wheel; but it's hard to deny that the vehicle is an enabling technology. Cars cause car accidents in the same way that guns kill people, which is after all why most societies impose regulations on both.

Google's lightning-quick access to the well of information too, it seems, can all but bring down an airline, although you still need a human idiot to punch a few keys to get you there. But is this really a new phenomenon? The rumour happened to be concocted via a lazy Google search and spread throughout the markets via the Web, but rumours existed before either the Internet or search engines.

True. But think for a moment about the pace of today’s events. In less than the time it takes to read this post, United Airlines shares lost a third of their value. Suddenly, that single piece of regurgitated memory from 2002 become today’s news, and it was everywhere. Almost eight hundred million dollars vanished, only to reappear in large part - and in different hands, one suspects - by day’s end. All because an old report resurfaced, popped up like a cork in the sea of information. Consider too that Google plans to extend its newspaper archives to the last 244 years, threatening to multiply the number of corks out there by several orders of magnitude. Perhaps the change is not one of degree, but something more radical.

Referring to Moore’s Law, which predicted in 1965 that the number of components on a chip would double every year, Stewart Brand has noted that ‘[a]ccording to a rule of thumb among engineers, any tenfold quantitative change is a qualitative change, a fundamentally new situation rather than a simple extrapolation’(2). At these speeds, the meaning of communication itself changes. And so too do its possibilities and dangers.

This weblog is about memory and technology. The term ‘technology’ has to be understood in the broadest possible sense - in a future instalment I shall argue for instance, and in very august company, that the alphabet itself is a technology. So most acts of committing to memory and indeed all acts of transmitting memory are technologically mediated by one or more languages, and a whole lot of apparatuses. The Internet is just another technology, in this regard, and it won’t be my sole concern; but it’s also one of the most topical and, let’s face it, interesting ones, as of right now. In discussing it, I shall indulge in many more stories like today’s, instances of dysfunction that undercut the experience that most of us have of using the Web and actually finding and exchanging all manner of useful information, without causing major financial collapses in the process.

There are two main reasons why I find these stories interesting. Scientists who study memory and the brain know that in order to figure out how we remember, it is useful to study individuals who can’t, be it as a result of a trauma or illness or genetic defect; similarly, I believe that developing memory practices commensurate to the sophistication of our communications technologies requires paying attention not only to the most exhilarating prospects on offer, but to the failures and the glitches, too. Secondly, and no less importantly, I believe that the anxiety and the concerns that these stories often touch upon or uncover are a most useful reminder of the material repercussions of our digital acts, and of the people and things and data that are left out, unable to remember or be remembered. A lot more on this, I’m afraid, in future posts.

One last thing: why the blog name, I hear nobody ask? I’m not going to go into that right now, except to credit the lovely artwork by Bert Warter, from the 1949 edition of Bruno Furst’s classic Stop Forgetting.

(1) Oliver James, 'My Mother Was Back. The Lights Were On', The Guardian, Saturday 2 August 2008.
(2) Stewart Brand, The Clock of the Long Now: Time and Responsibility (New York: Basic Books, 1999), p. 14.


stephen said...

There is another memory-related aspect to the United stock price story, which interests me as an investor.

"Even after the story was clarified and trading resumed, the airline shares still closed over 11 per cent lower than the day before."

I see that in the middle of trading a full day after, the price is STILL lower than Friday. Don't people remember Friday's price, and their assessment of value then? US financial markets may not be completely efficient, but they're certainly pretty efficient with wholly public information, so I find this very puzzling. Perhaps the memory of Monday's plunge, or the reawakened memory of 2002, have reduced people's optimism about United.

Giovanni said...

To what extent is that "assessment of value" rationally deduced on the basis of a cold analysis of the data, though? If it's rather a consensual rough estimate based on knowledge and expectation concerning past, present and future performance, plus all manner of irrational hunches I suspect, it would seem less suprising that bringing up the memory of a past trauma could interfere with that assessment, and that the effects would linger for a while.

As I say, I don't know much about investing. But I do derive some enjoyment from following the Nikkei historical trends, from time to time. Back in 1999, somebody told me I should really put my money (what money?) there, because the index had plummeted to 16,000 points from 39,000 points ten years earlier, and surely it was due back up - as we all know, in the long term shares always go up.

It's almost ten years later again and yesterday's close was 12,400.65.

stephen said...

In principle I believe in so-called value investing, which rests on the assumption that markets are not rational in short term but that they will be more so in the long term.

"In the short term the market is a voting machine; in the long term it is a weighing machine."

So no, I don't really believe that price and value are the same thing (though many people do), or that the price prevailing in the market is any more broadly correlated with a cold analysis of data. What we're seeing here is exactly the sort of phenomenon behavioural economics seeks to explain.

You might find the comments from "Mutant" and "amuseDetachment" in this discussion interesting. Briefly, the proposal is to use statistical analysis of online commentary on a market over some useful period to predict later price movements, assuming that there is a correlation. This can be seen as computer-aided arbitrage of the difference between stored memories (in computer data) and people's imperfect recollection in their own brain-stored memories.

Giovanni said...

I liked Mutant's complaint: "Why do participants depart from rational behavior, so often and to such extremes?" I've heard now and again people express the view that the capital markets would work seamlessly and predictably, if only people weren't involved. But of course you cannot un-involve people, since humans are enmeshed in the economic activity that underlies the stocks. And you cannot will the Nikkei back up towards 39,000 points either, just because every known historical trend demands it. So you're left with those pesky people doing their thing, and in that regard the idea of studying the chatter seems inspired. I'd love to get hold of the models.

As for stored memories vs. imperfect recollections, isn't that limiting in the assumption that the whole conversation takes place online? An imperfect brain might in fact recall a broader range of discussions, on and offline, and be a lot more successful in recalling the gist, the most salient snippets, as opposed to having to parse a whole lot of text that is meaningless (from the software's point of view) in search of word patterns and the like. But, as I say, I'd love to get hold of the models. And to be able to understand them.

Grunt said...

This is way too heavy for me to discuss with you. So I'm just going to post a little "happy blogging" comment and go away.