Tech and AI: 8. The Algorithm

This BBC podcast is supported by ads outside the UK.

So, what do this animal

and this animal

and this animal

have in common?

They all live on an organic valley farm.

Organic Valley dairy comes from small organic family farms that protect the land and the plants and animals that live on it from toxic pesticides, which leads to a thriving ecosystem and delicious, nutritious milk and cheese.

Learn more at ov.coop and taste the difference.

BBC Sounds, Music, Radio, Podcasts.

Welcome to Understand Tech and AI, the podcast that takes you back to basics to explain, explore, unpick, and demystify the technology that's becoming part of our everyday lives.

I'm Spencer Kelly from BBC Click and you can find all of these episodes on BBC Sounds.

YouTube, Google, Amazon, Facebook, Netflix, Twitter, Spotify, all of these services have so much stuff that they could show you.

Somehow they have to work out the stuff that you're going to be most interested in.

And the way that these services decide what to show you is they use artificial intelligence.

A specific form of it called the algorithm.

Can we get some echo on that?

No, right.

Never mind.

The algorithm seems to be all-powerful.

It seems to control what we see, what we're allowed to do.

And today we're going to try and understand it and if possible, outwit it.

And here to tell me whether I can do that is technology journalist Becca Caddy, who's also the author of Screen Time, How to Make Peace with Your Devices and Find Your Tequilibrium.

Becca, welcome.

Hi.

Do you think we are going to be able to outwit the algorithm in the next few minutes?

I think we can try.

I think we can get one step ahead at least.

Yeah.

Okay, so when we say the algorithm, what are we actually talking about?

There are many algorithms.

There is not a single algorithm.

They're incredibly complex, but if we're going to break it down to be as simple as possible, an algorithm is a set of instructions written in code that tells a computer how to solve a problem.

I often compare algorithms to recipes.

They're simple steps from A to B you do things in order and you get a result.

A lot of algorithms use AI and something called machine learning, which means they're no longer just following the steps, they're learning each time and making adjustments over time.

How do they work?

They're taking all kinds of information about you

and using that to then predict and decide what you'd like to see.

So whether that's a recommendation or an ad that you're more likely to click on.

Who makes these algorithms?

Computer scientists, engineers, software developers, people working for the big tech companies.

The whole point of them is to get you to stick around on the apps, on the site.

So they're central to how so many of the social media apps and things we use work.

And they're quite secret, aren't they?

Each company, like Facebook or Twitter or Netflix, they're quite secretive about how their algorithm works because that is their secret source, really, isn't it?

Yes, extremely secretive.

In fact, we don't know very much about exactly how any of them work.

Yeah, and they're all fighting for eyeballs.

They're all fighting for our time.

So it's very important that one service wins our attention over another.

Yes, exactly.

I mean, because on one hand, we don't have the time and energy to sort through everything, so we need help there.

But on the other, it is all to do with keeping our attention for as long as possible.

And I sometimes liken it to like a fun fair.

And if you just imagine that everything there, from the sweets to the rides, are just tailored to you, you're going to want to spend more time there.

So it kind of becomes no surprise that we can't tear ourselves away from Facebook or Instagram because it's intentionally everything we want to see.

And they can get really smart when it comes to looking at our habits, looking at the sorts of things that interest us.

So they can start to know us quite well, can't they?

Yeah, they really can.

So we're talking about algorithms using all sorts of information.

So simple things that you filled out on your profile, but also what do you look at?

What do you click on?

What do you like?

What do you just spend time looking on on your phone?

Even things like that are all playing back into learning more about you.

I've heard it said that these algorithms can know us better than we know ourselves.

Do you think that's true?

And or do you think they know us better than our friends and our family?

I think that's such an interesting question because, in some ways, I'm thinking of the way that the TikTok algorithm kind of knows if you linger on one video longer than another.

And you may not even be aware that you're doing that or you're interested in that thing.

So, there are metrics and things like that that I would say we're not consciously aware of.

And yet, at the same time, I do have to wonder: like,

something like that, as an example, what does that mean?

Like, does it mean I'm more interested in that thing?

Or did I just find that a little bit shocking and actually not like it, but just kind of lingered?

Because I was like, oh, I want to see how it ends.

But that is still showing that you want to watch that stuff, isn't it?

So it does give a little bit of use away to the algorithm.

Yeah, yeah.

So I'm not sure.

In some ways, absolutely, knows us better than ourselves.

How quickly can algorithms build up a picture of us?

I'm just going to give the really embarrassing example of the fact that I really like Pedro Pascal from The Last of Us.

Too embarrassing.

I love a bit of Pedro.

I may have searched for him on TikTok and then within minutes, all I'm seeing is Pedro Pascal-themed videos.

So, you know, I think it can act really, really, really fast.

Is there also a case where it will look at other people who are interested in Pedro Pascal and the sort of things they're also looking at?

And it will figure that you also might be interested in those other things yes absolutely because it happens in dating apps we know that it will happen elsewhere as well so the way dating apps work people are almost classified of a certain level of desirability and then people on the same level will be shown those people there are these kind of profiles that social media sites build can i just check have you been checking out pedro pasco on dating apps because i wish he was on dating you mentioned the two in the same sentence i'm just wondering

okay i think it's time we we pause for a second to cool off if nothing else.

Now, we've been talking about the algorithms that intelligently adapt to us and our habits, but the word algorithm is actually more broad than that.

So here's our resident tech historian, Dr.

James Sumner, to take us back, way back, to its origins.

The idea of an algorithm is as old as recorded history itself.

There are algorithms in the mathematical texts of ancient Babylon written over 4,000 years ago.

The roots of the name algorithm go back 1200 years.

It's a Latin version of Al-Khwaizmi, the name given to a Persian Muslim scholar whose writings were key to the spread of the decimal system from the Hindu and Arabic world into Europe.

Algorithms never needed machines.

They dictated the work of a lot of low-paid filing and calculating staff, but machines need algorithms.

Early efforts to explain computers to ordinary people were much more likely to talk about flow charts or demonstrate the concept with examples like sorting everyday objects.

Start at the right hand end, pick two up.

Yes.

The one in the right hand is lighter than the one in the left.

Swap.

Right, that's heavier.

Already on the first bus, you've got the lightest down, etc.

And we're starting all the way back to

the end.

And that's how a computer would do it.

That's exactly how a computer would do it, because the instructions are very simple.

Talk of algorithms only really hit the public consciousness around the 2010s, when people learned about the decision-making behind their Facebook feeds and a host of other interactions that set the tone of our online experience.

A string of scandals about data misuse raised awareness that these obscure mathematical structures actually hold tremendous political power and mistrust of algorithms as a potential threat flowed into other areas of everyday life.

In 2020, the mass cancellation of exam assessments in the UK during the COVID-19 pandemic led to an algorithmic attempt to simulate students' final grades, sparking public protests by school leavers, and they made it very clear where they placed the blame.

Grades that I received from my school were A star, A star, A, and I received A star, A star B by the algorithm, and because of that, I've lost out on my university place.

That was Dr.

James Sumner reminding us of the risks of leaving the algorithms to run things by themselves.

Okay, now, Becca.

There are also algorithms that make decisions about whether you're allowed certain things.

For example, mortgages.

How common are those?

Yeah, they're very common.

We see them in all kinds of places.

Like you say, mortgages.

There are algorithms that predict likelihood of disease and cancer risk and levels of care and things you might need.

So they are in a lot of different institutions.

This is something I think where people would be a bit fearful, a bit dubious, because it's a computer making a decision, not a human.

So it's dispassionate, it might get it wrong and we won't know why.

Are those reasonable fears?

So there have been a few cases now in the US of algorithms being used to determine healthcare.

And so there was one example, I think it was at Arkansas, of levels of at-home care that vulnerable people needed.

So these are people who had carers come in for maybe a certain number of hours per day.

And an algorithm was introduced to work out whether they really needed that level of care.

And there were just a lot of really awful repercussions.

Actually, a lot of people who had a lot of care throughout the day only had maybe an hour or maybe none at all.

And no one was really sure of exactly what had happened there.

And then it transpired that in the past, a human would have been there asking, I think it was around like 280, 300 questions of each person to determine levels of care.

But when the algorithm was brought in, I think there was only around 60.

So there is a huge part of these people's lives that just wasn't being considered all in the name of kind of streamlining.

And it took a long time in that particular case to work out what had even happened.

We hear examples of bias in algorithms.

Can you explain why an algorithm may end up being biased?

Algorithms aren't bad or evil, but if they haven't been properly tested or the data they've been trained on isn't right or, you know, again, the recipe analogy, if the algorithm doesn't know the difference between sugar and salt and it picks up salt to put in a cake recipe that's gonna make a huge mess so

um there have been a lot of examples of that actually and some of the worst is with facial recognition algorithms so there have been examples of algorithms that have been trained on a lot of data of white people's faces and not as much data on faces of black people now what that means is that the algorithm does a much worse job at discerning black faces So that's an error that could lead to all sorts of terrible repercussions.

The problem there is, you know, the algorithm was just working through the steps and the instructions, but it hadn't been given enough data.

I think the most important question of the episode has to be this.

How do we beat the algorithm?

I think we need to first manage our expectations.

We're not going to beat the algorithm.

What we can do is have more awareness that algorithms are everywhere.

And when it comes to our kind of own social media use,

I think really just being aware that everything we're seeing is curated.

And I think not only does that then give us a chance to tweak some settings on Instagram and Twitter, which are there.

So you're slightly less of maybe a slave to the algorithm.

But also I think having that awareness can really help.

I've found people who really struggle with habitual phone checking and think they may be addicted.

If they know that the algorithm is showing them things, like I said about an analogy of a fun fair, then that makes it easier to put your phone down as well.

You know what it's up to.

Yeah, you know what it's up to.

Don't blame yourself.

I mean, don't take it all on.

Oh, gosh, I can't control myself.

No, these things are deliberately trying to keep you looking at them.

Do we have to accept that the algorithm is here to stay and

can help us have a more interesting life as well as keeping us hooked?

I think we do.

Yeah.

Okay.

Unfortunately.

On that note, Becca, thank you for your time.

And Pedro Pascal, if you're listening, give us a call.

So what have we learned in this episode?

Artificial intelligence is watching our every move online.

It's learning our tastes and our habits.

And we're not seeing everything that's available.

We're just seeing a small part that's being curated by the algorithm.

Maybe being aware of that is at at least a good start if we want to keep control of our lives.

Mind you, if it's recommended this episode to you, well, thanks, AI.

Now, there's one question that keeps coming up whenever you mention artificial intelligence.

And in the next episode, we're going to have to face up to it:

Will AI

take my job?

That's next time.

For now, thanks for listening.

Suffs!

The new musical has made Tony award-winning history on Broadway.

We demand to be home.

Winner, best score.

We demand to be seen.

Winner, best book.

We demand to be qualified.

It's a theatrical masterpiece that's thrilling, inspiring, dazzlingly entertaining, and unquestionably the most emotionally stirring musical this season.

Suffs!

Playing the Orpheum Theater October 22nd through November 9th.

Tickets at BroadwaySF.com.