How do I get people to do what they're told?

If there’s one thing that would make everyone’s life easier right now, it’s other people doing what they’re told. 

But everyone – from world leaders to supermarket bosses to all of us with recalcitrant family members – is finding that go-to tactics like issuing instructions, posting lists, cajoling, pleading and screaming in frustration aren’t getting the job done.  

Would everything work better if the cyborg revolution speeded up and we got a bit more robotic about crisis-management? Or is this the time to double-down on our humanity? What follows are initial thoughts, do argue/add more in the comments.

How would a robot get my octogenarian Dad to self-isolate?

For all the hype about hyper-logic, AIs – in so far as thinking systems in silicon can be called intelligent – are no better at following instructions than humans. 

Like us, they listen selectively and do what meets their own priorities rather than thinking about the intent or even the goal behind the specific order.  Viktoria Krakovna compiled a great list of delinquent machine learning systems a couple of years back, which is fun reading even in these fearful times. Examples include an AI that falsified records to boost its score, several systems that exploit loopholes in game rules, others that subvert the whole point of the challenge by adopting absurd yet points-gaining strategies and – my personal favourite – systems that exploit lazy data-entry, such as the AI tasked with identifying deadly mushrooms that noticed that poisonous examples had been alternated consistently with the safely nutritious in the database, or the radiological AI that picked up on a consistent difference between pictures of non-cancerous and cancerous tumours: the yellow ruler included in pictures of the latter.

Then there’s speed and energy use. Deep learning systems, as I know to my cost, need to be trained in multiple iterations using vast data sets, which makes for big server costs and lots of exasperated conversations about how long it’s taking and whether we should just junk the tech and go back to basic statistical analysis on a desktop.  The same challenges come up time and again. The same solutions take just as long to implement. The same mistakes recur. And resist comprehensive solution.

Serious AI experts have sighed when I’ve brought up these examples and explained (wearily) that until with get AGI – artificial intelligence which can extrapolate and generalise rather than just processing through a bounded question – we have to be very, very precise about the challenges we set machine learning or other AI systems, and we have to iterate ruthlessly to reveal biases, loopholes, imprecisions and just plain junk that may have crept into our data and/or the models we are building.

I nod along to that line of reasoning…until I think back to humans.  General Intelligence is what we’re meant to be naturally good at, so how come the mistakes made by stupid – sorry, narrow – artificial intelligence feel so painfully close to home? And what can we all do to avoid the painful and frankly not very nice habit of issuing instructions, seeing them disobeyed or subverted, then muttering under our breath, as one boss of mine was wont to do, which part of “do that” did you not understand?

Atten-shun!

Yelling orders feels great. Such a shame it doesn’t get results.

The reason it’s so often counter productive is that humans don’t treat information neutrally. Our senses are primed not to take in the environment but to spot what’s important. That’s why our peripheral vision registers movement but not static objects; why Kim’s Game can reduce a birthday-partyful of children to howls of frustration; and why psychological experiments such as the watch-the-ball test produce such satisfyingly gotcha results. 

But, er, surely shouted commands count as important? 

Sadly not. Nothing like as important as the fact that the person who’s doing the shouting is pretty worked up. That – managing a powerful other’s emotional volatility – is what humans naturally focus on. Hence people who are shouted at typically retreat, get angry themselves, or go a bit numb – the flight/fight/lockdown natural reaction to a perceived threat to our pack membership and, hence, survival.

Like it or not, our stone-age brains are the best we’ve come up with, and have done pretty well by us so far, so rather than give up on influencing others, I thought I’d take a look at what big-data research has learned about what actually works when influencing people in times of crisis.

Data, schmata

It’s easy, when slogging through the psychological literature, to infer that big data and leadership effectiveness research are far-flung strangers. Too many so-called studies are actually anecdotes – stories of individual cases with little regard for comparability of methodology or data or anything approaching rigorous evaluation metrics. Others are simply hypotheses, often formed intuitively and, again, not tested across large populations (twenty people is only a large number when we’re talking about dinner guests, immediate family and people who would lend us a hundred quid at a moment’s notice). Then there’s a large group of findings that an American colleague of mine termed the Gee, Sherlock! methodology – in British idiom that translates to fancy ways of saying the bleedin’ obvious.

When we look at what’s left, the picture is even less rosy. Let’s take as an example the Leadership Styles developed by my former employer, HayGroup KornFerry. HGKF claims this methodology is robustly researched and evidence based, which is a bit of a dodge. It’s true that HGKF’s research into Organisational Climate – how the way it feels to work somewhere impacts results – is highly robust, having been tested repeatedly against outcomes in diverse conditions with large datasets. But the evidence for Leadership Styles’ influence on that Climate is less secure. The one big-data Leadership Styles project I was involved in, several years back, found that there was one just plain bad Style (essentially the screaming in people’s faces approach I discussed above) and that for all the other Styles it was horses for courses: certain Styles or combinations of Styles worked for certain people in certain situations. There was no magic formula for effective leadership or the influence of specific Styles on the genuinely results-boosting Climate.

So – you’ll be relieved to hear – I’m not going to drone you through anecdotes about different leadership approaches. Instead, let’s do what Sherlock, rather than Gee, Sherlock!, would do and, now that we’ve eliminated the impossible, follow what must – according to the data – be the truth.

I feel it in my fingers, I feel it in my toes…

The organisation that has probably amassed the most data about how people work is Google, which with Project Oxygen (what makes a great manager) and Project Aristotle (what makes a great team) has delved into the data on how people get other people to do stuff. In 2016 Project Aristotle reported that high performance was driven less by who was a on a team than by how the team worked together, particularly the degree of (in order of importance) psychological safety, dependability, clarity, meaning and impact team members feel.

This chimes with HGKF’s Organisational Climate research, which identified (in no particular order since I can’t access the original research!) Team commitment, Responsibility, Clarity,  Standards, Rewards and Flexibility.

A piece of big-data research I can access (because I was involved in it!) is another HGKF project, this one looking at Learning Climate within Primary and Secondary Schools in the UK – research paper posted on my LI profile, for the interested. Here again we found that Psychological Safety was key in helping children learn and adapt. For Primary Students (7-11 year olds) the strongest predictive factors were absolute levels of Order and Safety; for Secondary Students (11-18 year olds) particularly significant individual factors were Order, Interest, Clarity, Fairness and Safety.

Of course, Google employees are not precisely representative of the population, any more than children’s learning experience equates exactly to the experience of adults needing to learn new ways of living thanks to the threat of Covid19 infection. But these groups are not entirely unrepresentative, and the fact that similar themes – psychological safety, clarity, responsibility/order/fairness recur in all three big-data studies makes me think that we are likely on to something here.

I still don’t know what to say!

If I’ve learned anything in almost three decades of helping people and organisations change, it’s that it doesn’t matter what leaders say.

What matters is what people feel after they’ve heard you say it – the great-performance-producing feelings that Project Aristotle and the HGKF Climate research measures.

Focusing on your audience’s feelings rather than what you want to say doesn’t come naturally. Hands up if you’ve zoned out during a conference presentation, zoned back in, looked at the screen and seen a list of bullet points that are either too vague to make sense or too narrowly precise to be understood out of context. The problem isn’t your lack of attention, it’s that the presenter has forgotten the difference between aide-memoire and communication tool. A bullet point is a list, a summary of a richer argument or story. It makes sense only if you already know the story – don’t get this relationship backwards. Especially when the stakes are high – as NASA found out.

Three steps to start

Changing the way people work is never easy, changing our own ways of working is desperately hard. So don’t try to do it alone. If you’re in a role right now where you need to lead, where you need to help others feel safe, clear about what they need to do and secure in your ability to take care of them, get these kinds of support, ideally in the order below:

• Think about your own experience of feeling safer, clearer and more secure in the powers that be – what words or actions from leaders increased those feelings? Which messages resonated? What didn’t help (always useful to be reminded What Not To Do)?
• Borrow from the pros – look up examples that people cite when they talk about great communication and notice what’s said and how, as well as what’s not said. For starters check out the current NHS Covid19 factsheet (and contrast the official UK government action plan), and Churchill’s famous Blood, Sweat, Toil and Tears speech (bad examples too numerous to cite – trust me you’ll find them!)
• Test, test, test your approach with people who will give you honest and rich feedback on how they feel on hearing your words. Listen honestly even if it’s the last thing you expect or want to hear. Work the problem, calling in others for advice. Keep trying until a wide range of difficult-to-please people feel safe, clear and secure with your leadership.
• Keep on listening to the reaction. Even if you stick to the above points religiously, most likely your message won’t work at first attempt. But that doesn’t matter, so long as your first attempt doesn’t turn out to be your last attempt. I laughed bitterly in 2005 when London tubes were splattered with one of the most confusing and ungrammatical anti-terrorism warnings ever postered. But fast-forward to 2016, some learning from New York’s MTA, and the message had evolved to the simple and highly effective See it, Say it, Sorted.

Don’t ever forget that we are born to learn. And we get better together. 

Good luck, let me know how you get on and do continue the conversation in the comments.