Search This Blog

Showing posts with label random. Show all posts
Showing posts with label random. Show all posts

Sunday 1 September 2019

We know life is a game of chance, so why not draw lots to see who gets the job?

Sonia Sodha in The Guardian

Remove human bias from the interview process and the world might start to become a fairer place 


 
Interviews are an unreliable way of selecting the best person for the job. Photograph: Alamy


The sweaty palms, the swotting, the tricky question that prompts your heart to plummet: job interviews are no one’s idea of a good time. The other side of the equation is hardly fun either: days out of a busy schedule spent interviewing candidates, some of whom you know within a couple of minutes you would never offer a job.

Interviews are time-consuming for all involved. But we persist in doing them because recruitment decisions are some of the most important we take in the workplace and it follows we should invest time and energy into a robust recruitment process, right?

Wrong. It is long established that unstructured interviews are a notoriously unreliable way of selecting the best people for the job. This is perhaps unsurprising, when you consider the limited overlap between the skills needed to ace an interview and perform well day to day in a job or on a university course. And how many of us can honestly say we have been 100% truthful in a job interview?

Experimental studies show how unreliable interviewers are at accurately predicting someone’s capabilities. This is borne out on the rare occasions it gets tested in the real world. In the late 1970s, there was a doctor shortage in Texas and politicians instructed the state medical school to increase its admissions, after it had already selected 150 applicants after interview. So it took another 50 candidates who had reached the interview stage and been rejected, even though many of the stronger rejected candidates had already been snapped up by other medical schools. Researchers found these 50 students performed just as well as the original crop. Once the candidates got through the on-paper sift, they might as well have been drawn out of a hat.

Not only are interviews a generally bad way to spot talent, they are terrible at smuggling in bias. There are the obvious implicit biases – sexism, racism, ageism, class discrimination – but others also exist. According to psychologist Ron Friedman, we tend to perceive good-looking people to be more competent, tall candidates as having greater leadership potential and deep-voiced candidates as more trustworthy. Interviews also encourage us to pick people who look like us, think similarly to us and with whom we strike up an easy rapport. The myth of the meritocratic interview allows all sorts of prejudice to flourish.

These days, huge effort goes into trying to unpick these biases in interviews. Vast sums are spent on unconscious bias training, but the evidence as to its effectiveness is mixed at best. It turns out training a person’s subconscious to think differently isn’t as easy as a half-day course.


An element of random selection might engender a bit more humility on the part of white, middle-class men

This is why it is no substitute for breaking down the structures that allow these biases to fester. For example, managers might only be allowed to make an appointment once they have a sufficiently diverse shortlist. I’ve long been a believer in quotas for underrepresented groups where improving diversity is happening at a glacial pace, for example, in Oxbridge admissions.

But a recent conversation with a friend who works at Nesta, a charitable foundation, got me thinking about whether we should ditch the pretence that we can accurately predict people’s potential. Her organisation is experimenting with a lottery to award funding to staff for innovative projects. Employees can put forward their own proposal. All of those that meet a minimum set of criteria go into a draw, with a number selected for funding at random.

My initial thought was that this sounded bonkers. But ponder it more and the logic is sound. Not only does it eliminate human bias, it encourages creativity and avoids groupthink, discouraging staff from self-censoring because they think their idea is one management simply wouldn’t go for. It chimes with those who have argued that at least some science funding should be awarded by lottery, because in the contemporary world of peer review and scoring grids, risky ideas with potentially huge pay-offs do not attract sufficient funding.

Random selection embodies a very different conception of fairness to meritocracy. But if we accept that what we call meritocracy is predominantly a way for advantage to self-replicate, why not at least experiment with lotteries instead? Big graduate recruiters or Oxbridge courses could set “on paper” entry criteria, select candidates who meet them at random and test whether there are any differences with candidates selected by interview.

I am willing to bet that, as observed in Texas, they would do no worse. And that there would be other benefits: diversity of thought as well as diversity of demography. Quotas are often criticised for their potential to undermine those individuals who benefit from positive discrimination; everyone knows they are there not purely on merit, or so the argument goes. An element of random selection might engender a bit more humility on the part of white, middle-class men; it goes alongside being honest that meritocracy is a convenient mask for privilege.

The reason such experiments remain unlikely is that studies show that even when people are aware of the fallibility of interviews, they sustain incredible self-belief in their ability to buck the trend. Not only that, there are a lot of powerful people with a stake in maintaining the illusion of meritocracy. Oxford and Cambridge want to preserve the misconception that their selection procedures embody the creme de la creme of today selecting the creme de la creme of tomorrow.

But if you find yourself balking at random selection, ask yourself this: have you ever formed a first impression that was wrong? It might go against the grain, but making more liberal use of lotteries might produce not just a fairer but a better and more diverse world.

Friday 24 May 2019

It can be Hard to tell Luck from Judgment

Randomness often explains the difference between triumph and failure writes Tim Harford in The FT 


It hasn’t been a great couple of years for Neil Woodford — and it has been just as miserable for the people who have entrusted money to his investment funds. Mr Woodford was probably the most celebrated stockpicker in the UK, but recently his funds have been languishing. Piling on the woes, Morningstar, a rating agency, downgraded his flagship fund this week. What has happened to the darling of the investment community? 

Mr Woodford isn’t the only star to fade. Fund manager Anthony Bolton is an obvious parallel. He enjoyed almost three decades of superb performance, retired, then returned to blemish his record with a few miserable years investing in China. 

The story of triumph followed by disappointment is not limited to investment. Think of Arsène Wenger, for a few years the most brilliant manager in football, and then an eternal runner-up. Or all the bands who have struggled with “difficult second-album syndrome”. 

There is even a legend that athletes who appear on the cover of Sports Illustrated are doomed to suffer the “SI jinx”. The rise to the top is followed by the fall from grace. 

There are three broad explanations for these tragic career arcs. Our instinct is to blame the individual. We assume that Mr Woodford lost his touch and that Mr Wenger stopped learning. That is possible. Successful people can become overconfident, or isolated from feedback, or lazy. 

But an alternative possibility is that the world changed. Mr Wenger’s emphasis on diet, data and the global transfer market was once unusual, but when his rivals noticed and began to follow suit, his edge disappeared. In the investment world — and indeed, the business world more broadly — good ideas don’t work forever because the competition catches on. 

The third explanation is the least satisfying: that luck was at play. This seems implausible at first glance. Could luck alone have brought Mr Wenger three Premier League titles? Or that Mr Bolton was simply lucky for 28 years? Do we really live in such an impossibly random universe? 

Perhaps we do. Michael Blastland’s recent book, The Hidden Half, argues that much of the variation we see in the world around us is essentially mysterious. Mr Blastland’s opening example is the marmorkrebs, a kind of crayfish that reproduces parthenogenetically — that is, marmorkrebs lay eggs without mating and those eggs develop into clones of their mothers. 

Place two clones into two identical fish tanks and feed them identical food. These genetically identical creatures raised in apparently identical environments produce genetically identical offspring who nevertheless vary dramatically in their size, form, lifespan, fecundity, and behaviour. Sometimes things turn out very differently for no reason that we can discern. We might as well call that reason “luck” as anything else. 

This is not to say that skill doesn’t matter — merely that in a competition in which all the leaders are highly skilled, randomness may explain the difference between triumph and failure. Good luck plus skill beats bad luck plus skill any time. 

It is easy to underestimate how much chance is at play all around us. The psychologist Daniel Kahneman has recently been studying what he calls “noise”: the variability of judgments for no obvious reason. 

A wine expert blind-tasting two glasses from the same bottle will often rate them differently. Pathologists disagree with each other in their judgments of the same biopsy. More disconcertingly, they also disagree with their own prior judgments of the case. 

We rarely appreciate just how much inconsistency there is in the judgments we and others make, argues Prof Kahneman. It can hardly be a surprise, then, if past performance is no guarantee of future success. 

We should remember, too, that people often achieve outsized success by taking risks or being contrarian. When John Kay examined the forecasting record of economists in the 1990s, he noted that Patrick Minford, an idiosyncratic forecaster, would often produce the best forecast one year and the worst forecast the next. If the consensus is wrong, being an outlier gives you a high chance both of dramatic success and spectacular failure. 

We perceive all this randomness through a particular filter, too. Few people make the cover of Sports Illustrated after a run of mediocre luck. They appear after things have been going well, and if the good luck fails to hold then it seems like the SI jinx. More likely it is “regression to the mean”, or in simple terms, a return to business as usual. 

We begin paying attention only when someone is producing a remarkable performance. Genius followed by mediocrity is a story arc we all notice. Mediocrity followed by genius just looks like genius — assuming the mediocre performer gets a second chance. Not all do. 

So I wish Mr Woodford well. Perhaps he has lost his touch, perhaps the world has changed, or perhaps he has simply been unlucky. It would be nice to know which, but in such matters the world does not always satisfy our curiosity.

Saturday 22 October 2016

So much for scientific publications: Nonsense paper written by iOS autocomplete accepted for conference

Elle Hunt in The Guardian


A nonsensical academic paper on nuclear physics written only by iOS autocomplete has been accepted for a scientific conference.


Christoph Bartneck, an associate professor at the Human Interface Technology laboratory at the University of Canterbury in New Zealand, received an email inviting him to submit a paper to the International Conference on Atomic and Nuclear Physics in the US in November.

Since I have practically no knowledge of nuclear physics I resorted to iOS autocomplete function to help me writing the paper,” he wrote in a blog post on Thursday. “I started a sentence with ‘atomic’ or ‘nuclear’ and then randomly hit the autocomplete suggestions.

“The atoms of a better universe will have the right for the same as you are the way we shall have to be a great place for a great time to enjoy the day you are a wonderful person to your great time to take the fun and take a great time and enjoy the great day you will be a wonderful time for your parents and kids,” is a sample sentence from the abstract.

It concludes: “Power is not a great place for a good time.”
Bartneck illustrated the paper – titled, again through autocorrect, “Atomic Energy will have been made available to a single source” – with the first graphic on the Wikipedia entry for nuclear physics.

He submitted it under a fake identity: associate professor Iris Pear of the US, whose experience in atomic and nuclear physics was outlined in a biography using contradictory gender pronouns.

The nonsensical paper was accepted only three hours later, in an email asking Bartneck to confirm his slot for the “oral presentation” at the international conference.

“I know that iOS is a pretty good software, but reaching tenure has never been this close,” Bartneck commented in the blog post.

He did not have to pay money to submit the paper, but the acceptance letter referred him to register for the conference at a cost of US$1099 (also able to be paid in euros or pounds) as an academic speaker.

“I did not complete this step since my university would certainly object to me wasting money this way,” Bartneck told Guardian Australia. “... My impression is that this is not a particularly good conference.”

The International Conference on Atomic and Nuclear Physics will be held on 17-18 November in Atlanta, Georgia, and is organised by ConferenceSeries: “an amalgamation of Open Access Publications and worldwide international science conferences and events”, established in 2007.

An organiser has been contacted by Guardian Australia for comment.

Bartneck said that given the quality of the review process and the steep registration fee, he was “reasonably certain that this is a money-making conference with little to no commitment to science.

“I did not yet reply to their email, but I am tempted to ask them about the reviewers’ comments. That might be a funny one.”

The conference’s call for abstracts makes only a little more sense than Bartneck’s paper.

“Nuclear and sub-atomic material science it the investigation of the properties, flow and collaborations of the essential (however not major) building pieces of matter.”

A bogus research paper reading only “Get me off Your Fucking Mailing List” repeated over and over again was accepted by the International Journal of Advanced Computer Technology, an open-access academic journal, in November 2014.

Friday 2 January 2015

We can’t control how we’ll die. The limits of individual responsibility


It’s important to live healthily, but scientists also tell us that the majority of cancers are down to chance – a good reminder of the limits of individual responsibility
Close-up of two stubbed-out cigarettes
'Most cancers are a random lightning bolt, not something we can avoid by keeping away from tobacco or excessive booze or by going for regular morning runs.' Photograph: Alamy

Our terror of death (happy new year, by the way) surely has much to do with a fear that it is out of our control. The lifetime risk of dying in a road accident is disturbingly high – one in 240 – and yet the freakishly small chance of dying in a plane accident generally provokes far more fear. We dread a final few moments in which we are powerless to do anything except wait for oblivion. So perhaps the news that most cancers are the product of bad luck – rather than, say, our diet or lifestyles – is scant reassurance. Most cancers are a random lightning bolt, not something we can avoid by keeping away from tobacco or excessive booze, or by going for regular morning runs. That’s something we have to live with.
But perhaps the news should be of comfort. It is, of course, crucial to promote healthy lifestyles. Regular exercise, a good diet and the avoidance of excess does save lives. Yet the cult of individualism fuels the idea that we are invariably personally responsible for the situation we are in: whether that be poverty, unemployment or ill health. Cancer is more individualised than most diseases: all that talk of “losing” or “winning” battles. A far wiser approach was summed up by DJ Danny Baker after his own diagnosis. He said he was “just the battlefield, science is doing the fighting and of course the wonderful docs and nurses of the brilliant NHS”. The cancer patient, in other words, was practically a bystander in a collective effort.
One of the heroes of 2014 was Stephen Sutton because of his infectious optimism and cheerfulness in the face of cancer. But his battle was about not letting cancer consume his final few months on earth, rather than a superhuman quest to miraculously defeat the disease himself. What struck me about Stephen was that a situation that seemed nightmarish to most of us became an opportunity for him to take control of his life. It is what struck me, too, about Gordon Aikman, a 29-year-old Scot with a terminal diagnosis of motor neurone disease. There is no right way to die, but he has learned how to live.
So that’s why I have some sympathy with Richard Smith, a doctor who once edited the British Medical Journal. He has upset many by suggesting we are “wasting billions trying to cure cancer”, when it is the “best” way to go. I certainly would not advocate cutting back on cancer research, quite the opposite – even if other fatal diseases don’t receive the same amount of attention – and cancer can be a horrible way to die. But his point was that it provided an opportunity to make peace, to reflect on life, to do all the things you always wanted to do – to finally have control over your own life. Other ways of dying simply do not provide that option, either because they are so sudden or because of the form they take.
We have less of a say over how and when we die than we thought. That may be a cause for anxiety: it may actually frighten us more. I think it’s liberating. If only we learned to live like many of those – like Stephen or Gordon – facing death, taking control of their lives, we would be so much happier than we are.

Sunday 20 October 2013

Nobel Prize winners say markets are irrational, yet efficient

S A Aiyar in The Times of India
Are stock markets irrational, driven by greed and fear, subject to euphoria and panic? Or are they highly efficient indicators of intrinsic value? Both, says the Nobel Prize Comittee for Economics, with no sense of contradiction.
It has just awarded the prize jointly to economists with opposing views. Robert Shiller is famous for two versions of his book 'Irrational Exuberance'. The first version appeared in 2000 at the height of the dotcom boom, and correctly predicted that this was a bubble about to burst. The second version came in 2005 just as the housing market was skyrocketing, and predicted (again correctly) that this too was a bubble likely to burst resoundingly.
This confirmed Shiller's status as a behavioural economist. Such economists laugh at the notion that human beings are rational economic actors, as portrayed in textbooks. No, say behaviourists, humans are driven by fads, prejudices, manias, and irrational bouts of optimism and pessimism. Yet Shiller is going to share the Nobel Prize with Eugene Fama, famous for his "efficient markets hypothesis." This states that markets are like computers processing information from millions of sources on millions of economic actors, and hence produce more efficient long-run valuations than the most talented genius.
Fama's market behaviour is fundamentally random, so future trends cannot be predicted by even the cleverest investors. He implies that choosing stocks by throwing darts at a stock market chart can beat the recommendations of top experts. This has been verified by some, though not all, dartthrowing contests.
Corollary: ordinary investors must not pay high fees to experts to pick winners. Instead they should invest passively in a group of shares (like the 30 shares constituting the Bombay Sensex or Dow Jones Industrial Average), and ride these bandwagons without paying any fees. This has led to the spectacularly successful emergence of Index-traded funds (like those run by Vanguard in the US). Such funds are indexed to share groups like the Sensex or the Banks Nifty. Rather than try to pick individual winners in say the auto, pharma or realty sectors, index funds invest passively in a group of auto, pharma or realty companies. This has proved successful and popular.
Two groups criticize the efficient markets hypothesis: big investment gurus and, paradoxically, leftists viewing financial markets as instruments of the devil. Investment gurus like Warren Buffett in the US or Rakesh Jhunjhunwala in India claim to have beaten the market average handsomely, thus disproving the efficient markets hypothesis. Not so says Fama: in any large collection of investors there will always be some who perform above average and some below average - this is a matter of statistical chance, not skill. Moreover, investment gurus have so many contacts that they may have insider information enabling them to beat the market by unfair means.
As for Shiller's successful predictions, Fama says capitalism is driven by booms and busts. To predict at the height of a boom (like Shiller) that a bust will follow is banality, not genius. It is as unremarkable to predict during every bust that a boom will follow.
After the 2008 global financial crisis, the new conventional wisdom is that governments need macro prudential policies to check future financial crises, and that finance should be more strictly regulated than ever before. However, the counter is that the financial crisis occurred even though the financial sector was already the most regulated (with 12,000 regulators in the US alone). Governments had encouraged reckless lending by guaranteeing large banks and investment banks against failure, and by creating governmentbacked underwriters like Fannie Mae who shouldered any burdens caused by mass default.
Perhaps the Nobel Prize Committee is right in implying that markets can be both irrational and efficient at the same time. Since humans are irrational, they will always create markets that have booms and busts, marked by irrational optimism and pessimism. An efficient markets defined by Fama and his followers is not one that produces steady growth without booms, busts or crises. It is efficient only in the limited sense that, whether the markets are calm or irrational, they represent the processed information of millions of actions of millions of actors, and this is inherently more efficient than the efforts of any individual investor.
The argument is analogous to the one against communism or dictatorship. Communists believed that the great and good politburo, motivated entirely by the public interest and not profit, would run the economy better than the chaos, irrationality and imperfections of the capitalist market. Yet the market, with all its flaws and irrationality, proved infinitely more efficient.
Fama holds that this is true of financial markets too. This is compatible with Shiller's analysis. Markets can be both irrational and efficient.

Wednesday 1 May 2013

Cricket and Causes - It's not about selection or tactics, silly


Understanding causes is incredibly difficult. It is much easier to assume that easily discernible surface issues are the primary explanations for victory and defeat
Ed Smith
May 1, 2013



England v Australia, The Ashes 3rd npower Test, Nottingham, 02-06 Aug 2001
Mike Atherton copped criticism during Australia's dominance in the 1990s: "It is not easy to be bold, consistent or whatever else is deemed topical, when you are losing matches" Paul McGregor / © ESPNcricinfo Ltd 
Enlarge
Related Links
If you want to understand sport, you have to understand causes. More accurately, you have to understand how difficult it is to be sure about which causes really influence events, and which are merely irrelevant side issues.
Coaching is about understanding causes: what causes players to perform better? Journalism is about causes: which factors led one team to beat the other? Fans, too, reflect obsessively about causes: what might make the difference for us next season? Sport, like history, is about causes.
And yet understanding causes is incredibly difficult. Causal threads must be observed and disentangled, then weighed and judged. It is much easier simply to assume that easily discernible surface issues - such as selection and short-term tactics - are the primary explanations for why teams win and lose.
That is why the books that have most influenced my thinking about sport address the question of causes rather than sport itself. If I had to name one book that anyone with a serious interest in sport should read, it would be Nassim Taleb's Fooled by Randomness. It scarcely mentions sport, and Taleb actively dislikes organised games. But Fooled by Randomness explores the dangers of sloppy assumptions about causality. It attacks lazy guesses about one thing "leading" to another. It makes the reader re-examine his own flawed reasoning.
Taleb recalls watching the financial markets on Bloomberg TV in December 2003. When Saddam Hussein was captured, the price of US treasury bills went up. The caption on TV explained that this price movement was "due to the capture of Saddam Hussein". Half an hour later, the price of US treasury bills went down. The TV caption explained that this was "due to the capture of Saddam Hussein".
The same "cause" had been invoked to "explain" two opposite effects, which is, obviously, logically impossible.
The next time you absorb sports punditry, keep in mind that story about Bloomberg TV and the price of Treasury bills. You will learn that a golfer misses a crucial putt "because he lost concentration", and then misses the next putt because he was "trying too hard". You will learn that a team loses one match "because they didn't stick to the game plan", then loses the next "because they were unable to think on their feet".
A manager messes up one match "because he was too loyal to his favourite players", then fails in the next "because he unnecessarily alienated the core of the team". And, my favourite: there is always the player who "benefits from utter single-mindedness" one week, and then "suffers from a damaging lack of perspective" the next.
The point, of course, is that causes are being manipulated to fit outcomes. They weren't causes at all, merely things that happened before the defeat. The ancient Romans had an ironic phrase for this terrible logic - post hoc, ergo propter hoc, "after this, therefore because of this".
It is hard to imagine a stronger contender for adopting false causes than the failure of English cricket teams to win the Ashes between 1987 and 2005. This dismal sequence was, apparently, "caused" by the following factors: structure of county cricket, unshaven stubbles worn by some England captains, sticking with a failing core of senior players for too long, introducing too many new players, being insufficiently hard-working and professional, being insufficiently joyful and amateur, having too many counties, being too English, not being English enough. And so on.
Pretty much anything that existed within English cricket, at some point or other, was used to explain England's lack of success in the Ashes. An English cricketer in the 1990s only had to brush his teeth to be told that they didn't do it like that in Australia.
Above all, English cricket failed because it was not like Australian cricket. If only England teams would copy Australian teams by (in no particular order): swearing/caring/sledging/bonding/singing/ drinking/attacking/being mates/taking risks/backing themselves/fronting up/digging in/manning up/playing for the badge/never saying die… if England teams simply did all that, then, frankly, playing Shane Warne's flipper and Glenn McGrath's metronomic seam-up would be a doddle.
When your best is not quite good enough, the two levers under your control - selection and tactics - begin to look very inadequate. In other words, they are not really "causes" of defeat at all. They are simply things that happened along the way
Imagine the logical gymnastics required when England started winning Ashes series again. All the previous causes of defeat had now to be converted into explanations for victory. If England's Ashes success continues, it can only be a matter of time until we have the ultimate "Bloomberg moment", when an article is written arguing that Australia routinely loses the Ashes because they have too few state sides and must urgently copy England's first-class structure of 18 counties.
True, some things within English cricket have changed in reality as well as perception: players are now centrally contracted to the England team, for example, rather than to their counties. But not as much has changed as is often claimed. Revolution - "chumps to champs" - is a snappier narrative than gradual evolution.
But the real fun lies elsewhere. It has now become fashionable to scour Australian cricket looking for "causes" of their decline. A few years ago, the personality of Michael Clarke became the focal point for critics of the culture within Australian cricket. When Clarke came good, it was time to look elsewhere for "causes" of muted Australian performances. Ex-players attacked selection as confused, even insulting. Australia, they argued, had to pick more young players, and yet had to pick more players with hard-earned experience; they had to stick with a consistent team while also, inevitably, abandoning obvious mistakes. Sound familiar?
Mike Atherton, the former England captain who received his fair share of criticism during the era of Australian dominance, remarked wryly this week: "It is not quite so easy to be bold, to be consistent or whatever else is deemed topical, when you are losing matches."
The two central variables in sport, the main levers controlled by the management, are selection and tactics. Imagine, for a moment, that you are in charge of the lesser of two teams. You pick what you think is your best XI. And you lose, despite the team playing at or near its potential. If you stick with the same team, are you not merely sleepwalking towards another defeat? And yet if you change it, what has led you to change your mind about the team that you thought was the best XI last week and which, after all, did not really under-perform? Difficult one, isn't it, picking a team that is less good than the opposition?
Now tactics. Imagine you devise what you consider to be your optimal tactical approach. You execute the plan reasonably well. And you lose. Do you change tactics, with the same logic that led you to change the team, or stick with the old tactics that led to defeat?
Very simply, when your best is not quite good enough, the two levers under your control - selection and tactics - begin to look very inadequate. In other words, they are not really "causes" of defeat at all. They are simply things that happened along the way.
It is the same with national economics. Governments and central banks control the familiar levers of interest rates, money supply and taxation. They are endlessly criticised for their handling of all three. But what if the actual economy, the thing itself, is simply not very robust? A rabbit cannot always be conjured magically from a hat.
I would not have explored all this if I wasn't surprised at how often it is forgotten or overlooked in the analysis of sport at every level, from the pub to the board room, and from the commentary box to the armchair. We have long accepted that understanding historical causes is profoundly subtle and intellectually demanding. Exactly the same applies to understanding causes in sport.