1
00:00:00,000 --> 00:00:19,120
[Music]

2
00:00:19,120 --> 00:00:21,840
Welcome back to Quietly Secure.

3
00:00:21,840 --> 00:00:26,160
Over the last few episodes, we've been talking about trust.

4
00:00:26,160 --> 00:00:32,560
Why we trust websites, why scams work, and how artificial intelligence can create

5
00:00:32,560 --> 00:00:39,440
convincing images, voices and videos, and whether we can trust the answers that AI gives us.

6
00:00:39,440 --> 00:00:44,800
There's a common thread running through all of these subjects, and that's information.

7
00:00:44,800 --> 00:00:51,280
More specifically, how do we know whether the information we're receiving is worth trusting?

8
00:00:52,480 --> 00:00:56,880
For most of human history, finding information was relatively difficult.

9
00:00:56,880 --> 00:01:04,080
You might buy a paper, watch the evening news, listen to the radio, read a book,

10
00:01:04,080 --> 00:01:10,320
speak to someone who knew something you didn't. Today, information is everywhere,

11
00:01:10,320 --> 00:01:15,760
almost too much of it, and the challenge isn't finding information anymore,

12
00:01:15,760 --> 00:01:22,240
it's deciding what deserves our attention, because the internet has changed something fundamental.

13
00:01:22,880 --> 00:01:28,160
It doesn't just give us information, it decides what information we see,

14
00:01:28,160 --> 00:01:31,520
and increasingly, it does that through algorithms.

15
00:01:31,520 --> 00:01:39,280
Think about traditional newspaper, a group of editors decided what was important.

16
00:01:39,280 --> 00:01:45,840
They choose the stories, they decide the order, they write the headlines,

17
00:01:45,840 --> 00:01:49,040
and eventually the newspaper arrives at your doorstep.

18
00:01:49,680 --> 00:01:55,280
You didn't choose every story, you received the selection someone else had made.

19
00:01:55,280 --> 00:02:02,080
Television news worked in much the same way, there was a limited amount of time,

20
00:02:02,080 --> 00:02:06,560
editors had to decide what mattered, then came the internet.

21
00:02:06,560 --> 00:02:12,080
Suddenly, there was no particular limit on how much information could be published.

22
00:02:12,080 --> 00:02:18,960
Thousands of articles, millions of posts, videos, podcasts, comments,

23
00:02:18,960 --> 00:02:25,680
social media updates, the problem changed. We know a longer needed help finding information,

24
00:02:25,680 --> 00:02:30,880
we needed help filter in it, and that's where algorithms come in.

25
00:02:30,880 --> 00:02:39,600
The word can sound complicated, but an algorithm is essentially a set of instructions for solving

26
00:02:39,600 --> 00:02:46,800
a problem or making a decision. In the world of online services, algorithms help decide what

27
00:02:46,800 --> 00:02:52,960
you're shown. Which video appears next? What posts appear near the top of your feed?

28
00:02:52,960 --> 00:03:00,400
Which search results appear first? Which products are recommended? And which stories are you

29
00:03:00,400 --> 00:03:06,800
likely to click? The important thing is that these systems aren't usually trying to answer

30
00:03:06,800 --> 00:03:11,920
one simple question, they're trying to predict what you're likely to do.

31
00:03:13,040 --> 00:03:18,720
Will you click? Will you watch? Will you read? Will you share? Will you come back tomorrow?

32
00:03:18,720 --> 00:03:25,440
And that creates an interesting situation. The information that gets your attention

33
00:03:25,440 --> 00:03:29,120
isn't necessarily the information that's most important.

34
00:03:29,120 --> 00:03:37,200
Your attention has become an extremely valuable resource. If you spend an extra 10 minutes

35
00:03:37,200 --> 00:03:43,440
watching videos, that's valuable to a platform. If you click on another article,

36
00:03:43,440 --> 00:03:49,840
that's another opportunity for advertising. And if you keep scrolling, you remain engaged.

37
00:03:49,840 --> 00:03:57,200
This doesn't automatically make algorithms bad. Recommendation systems can be incredibly useful.

38
00:03:57,200 --> 00:04:03,680
The helpers find music will like, films where you might enjoy, products where interested in,

39
00:04:04,320 --> 00:04:11,600
information relevant to our work. The problem is, the system doesn't necessarily know what is good.

40
00:04:11,600 --> 00:04:16,320
It knows what is effective. And those aren't always the same thing.

41
00:04:16,320 --> 00:04:23,120
Consider something that makes you angry. You see a headline, you disagree with it,

42
00:04:23,120 --> 00:04:29,200
you immediately want to respond. Perhaps you share it. Maybe you write a comment,

43
00:04:29,200 --> 00:04:34,960
you return to see what other people have said. And from that platform's perspective,

44
00:04:34,960 --> 00:04:42,880
you've just become highly engaged. Your anger generated activity, and activity is measurable.

45
00:04:42,880 --> 00:04:48,480
That doesn't mean every controversial story is deliberately designed to make you angry.

46
00:04:48,480 --> 00:04:55,920
But it does mean emotional content can perform extremely well. An algorithms learn from performance.

47
00:04:56,560 --> 00:05:02,080
If something consistently gets attention, the system has a reason to show more of it.

48
00:05:02,080 --> 00:05:10,080
There's another effect. The more you interact with certain types of information,

49
00:05:10,080 --> 00:05:16,880
the more likely you are to receive similar information. If you regularly watch technology videos,

50
00:05:16,880 --> 00:05:23,200
you're probably going to see more technology videos. If you read about a particular political

51
00:05:23,200 --> 00:05:29,760
subject, you'll probably receive more related material. If you repeatedly interact with a

52
00:05:29,760 --> 00:05:36,400
particular viewpoint, the system learns that this is something you engage with. Eventually,

53
00:05:36,400 --> 00:05:42,320
the internet you experience can become very different from the internet experience by somebody

54
00:05:42,320 --> 00:05:48,800
sitting next to you. You can both open the same platform and see completely different worlds.

55
00:05:49,440 --> 00:05:56,960
This is an important idea. There isn't necessarily one internet anymore. There are personalized

56
00:05:56,960 --> 00:06:04,560
versions of it. And this is where things can become unnecessarily dramatic. Algorithms aren't

57
00:06:04,560 --> 00:06:11,040
necessarily sitting somewhere, deciding what you should believe. Most of the time, they're doing

58
00:06:11,040 --> 00:06:18,720
something much simpler. They're optimizing a particular objective. More engagement, more viewing,

59
00:06:18,720 --> 00:06:26,000
more interaction, more retention. The unintended consequences that the information that generate

60
00:06:26,000 --> 00:06:32,560
strong reactions can sometimes outperform information that is accurate but boring.

61
00:06:32,560 --> 00:06:40,480
A careful research article might take 10 minutes to read. A sensational headline can get a

62
00:06:40,480 --> 00:06:47,840
reaction in three seconds. The algorithm can measure those three seconds. It can't necessarily

63
00:06:47,840 --> 00:06:55,680
measure whether the article improved your understanding of the world. But the answer isn't to stop

64
00:06:55,680 --> 00:07:04,000
using social media. Nor is it to distrust everything we see. Instead, be aware of the system.

65
00:07:04,000 --> 00:07:11,440
If something makes you extremely angry, pause. If something makes you incredibly excited, pause.

66
00:07:12,240 --> 00:07:19,760
If something confirms exactly what you already believe, perhaps pause again. Not because it's

67
00:07:19,760 --> 00:07:28,160
necessarily false, but because emotional certainty is a good reason to check. Look for another source,

68
00:07:28,160 --> 00:07:36,640
read beyond the headlines and consider the date. Find the original information. And perhaps most

69
00:07:36,640 --> 00:07:43,840
importantly, don't confuse popularity with truth. A million views tells you that a million people

70
00:07:43,840 --> 00:07:51,200
watch something. They don't tell you whether it was accurate. One of the most useful digital skills

71
00:07:51,200 --> 00:07:58,080
isn't technical at all. It's being comfortable saying "I don't know". We don't have to have an

72
00:07:58,080 --> 00:08:05,280
opinion about everything. We don't have to react to every story. We don't have to share something

73
00:08:05,280 --> 00:08:11,360
simply because everyone else is talking about it. Sometimes the most secure thing you can do

74
00:08:11,360 --> 00:08:20,080
online is nothing. Let that story sit. Wait for more information. Come back tomorrow.

75
00:08:20,080 --> 00:08:26,480
Quite often the world becomes a little clearer when the initial emotion has passed.

76
00:08:26,480 --> 00:08:33,600
The internet hasn't simply changed how we received news. It's changed how information

77
00:08:33,600 --> 00:08:39,920
competes for our attention. Algorithms help us navigate an overwhelming amount of content.

78
00:08:39,920 --> 00:08:46,400
But they're not necessarily designed to tell us what matters. They're designed to predict what

79
00:08:46,400 --> 00:08:53,600
will do. Understanding that distinction changes how we look at our feeds. The next time something

80
00:08:53,600 --> 00:08:59,040
appears on your screen asks yourself "why am I seeing this?" Not just is it true.

81
00:08:59,840 --> 00:09:05,600
The first question can sometimes be just as valuable because understanding the system around

82
00:09:05,600 --> 00:09:12,720
the information can help us understand the information itself. Thank you for listening to Quietly

83
00:09:12,720 --> 00:09:19,760
Secure. Next time we'll turn our attention back towards ourselves. Because while we're busy looking

84
00:09:19,760 --> 00:09:27,120
at everyone else's lives online, other people are forming opinions about us. And increasingly,

85
00:09:27,120 --> 00:09:32,960
so are the systems behind the screen. We'll be talking about digital reputation.

86
00:09:32,960 --> 00:09:52,240
Who are you online? Until then, stay curious, stay calm, and as always, stay Quietly Secure.

87
00:09:52,240 --> 00:10:02,920
[END PLAYBACK]