Social Media Algorithms and How They Rule Our Lives

We all know how social media algorithms work, right?

Most people think they know how social media algorithms work, but they don’t. This art­icle will shed some light on how algorithms use iter­at­ive test­ing — not soph­ist­ic­ated technology.

As a digit­al strategist, I’ve closely mon­itored social media algorithms for two dec­ades. I’ve learnt that algorithms aren’t your friends — and we must man­age them.

Here we go:

We Need Sorting and Labelling

You inter­act with social media, and the plat­form own­er col­lects your user data to serve you more con­tent to keep you engaged and thus increase your expos­ure to third-party advertising. 

“So what’s the Original Sin of the Internet? Nearly all busi­ness mod­els it sup­ports require spy­ing on con­sumers and mon­et­ising them.”
— Bob Sullivan, author and journalist 

Order is neces­sary, of course, since there’s a lot of con­tent to struc­ture:

Unfortunately, the actu­al inner work­ings of a social media algorithm have a much dark­er side. And yes, “dark­ness” is a reas­on­able ana­logy because these algorithms are being kept secret for many reasons.

Behind these cur­tains of secrecy, we don’t find myri­ads of com­plex com­put­ing lay­ers but man­u­fac­tured fil­ters designed by real people with per­son­al agendas. 

We are con­stantly shap­ing, sug­gest­ing, nudging, and presenting.

Social Algorithms are Not “Personal”

As users, we have a gen­er­al idea of how algorithms work, but only a hand­ful of people know. To pre­vent indus­tri­al espi­on­age, we can safely assume that most social net­works are mak­ing sure that no one developer has full access to the entirety of an algorithm.

And even if you’re a Facebook pro­gram­mer, how would you know exactly how Google’s algorithm works?

One might assume you have a per­son­al Facebook algorithm stored on a serv­er some­where. An algorithm that tracks you per­son­ally and learns about you and your beha­viour. And the more it knows about you, the bet­ter it under­stands you. But this is not exactly how it works — for a good reason.

Humans are notori­ously bad at con­sciously know­ing ourselves and under­stand­ing oth­ers. And our think­ing is riddled with uncon­scious biases. 

However com­plex, apply­ing vari­ous types of machine learn­ing to learn about users and their inter­ac­tions on the indi­vidu­al level would be both slow and expens­ive. Few social media users would be patient enough to endure such a lengthy pro­cess through tri­al and error.

Anyone famil­i­ar with data min­ing will know that more advanced scrap­ing tech­niques, like sen­ti­ment ana­lys­is from social media mon­it­or­ing, will require large data sets. Hence, the term “big data.” ¹Batrinca, B., Treleaven, P. Social media ana­lyt­ics: a sur­vey of tech­niques, tools and plat­forms. AI & Soc, 89 – 116 (2015). https://doi.org/10.1007/s00146-014‑0549‑4

A social media algorithm gets immense power primar­ily from har­vest­ing data from large users sim­ul­tan­eously and over time, not from cre­at­ing bil­lions of self-con­tained algorithms. 

Put anoth­er way: Algorithms are fig­ur­ing out human­ity at scale, not your behaviour.

Our Lack of Consistency

No one argues the fact that you are an indi­vidu­al. But is your beha­viour con­sist­ent from inter­ac­tion to inter­ac­tion? The answer is prob­ably no. How you act and react will likely be much more con­tex­tu­al and situ­ation­al — at least con­cern­ing your per­ceived unique­ness and self-identification.

Social net­works are lim­it­ing your options to tweak “your algorithm” your­self. I would be all over the pos­sib­il­ity of adjust­ing Facebook’s news­feed or Google’s search res­ults in fine detail using vari­ous boolean rule­sets. However, that would prob­ably only teach the mas­ter social algorithm more about human pre­ten­sions — and little else.

Accurate per­son­al algorithms that fol­low us from ser­vice to ser­vice could out­per­form all oth­er algorithms. I would prob­ably ask my algorithm only to show me ser­i­ous art­icles writ­ten in peer-reviewed pub­lic­a­tions or pub­lished by well-edu­cated authors with proven track records. But if the algorithm does­n’t take it upon itself to show me some funny cat memes or weird YouTube clips now and then, I would prob­ably be bored quite quickly.

When we under­stand that the social media algorithms aren’t try­ing to fig­ure you out, it fol­lows to ask how well they’re doing in fig­ur­ing out humanity.

Why Social Media Algorithms Aren’t Better

Social media algorithms are pro­gress­ing — painstak­ingly. Understanding human beha­viour at the macro level is not a task to be underestimated.

The prac­tic­al approach to these struggles is straight­for­ward:
Eliminate the guesswork.

Algorithms: The Silent Switch

Not that long ago, social media algorithms would deliv­er organ­ic reach accord­ing to a dis­tri­bu­tion that looked like this:

Today, social media algorithms deliv­er organ­ic reach more like this:

It’s the Silent Switch where social net­works have demoted the pub­lish­er­’s author­ity and repu­ta­tion and pro­moted single con­tent per­form­ance instead.

This algorithmic change has likely had pro­found and severe media logic amplifications:

Classic Media Logic Effects

Classic media logic is hypo­thes­ised to influ­ence the news media in the fol­low­ing ways: ²Nord, L., & Strömbäck, J. (2002, January). Tio dagar som skakade världen. En stud­ie av medi­ernas beskrivningar av ter­ror­at­tack­erna mot USA och kri­get i Afghanistan hösten 2001. ResearchGate; … Continue read­ing

Social Media Logic Effects

Our job as PR pro­fes­sion­als is to help organ­isa­tions nav­ig­ate the media land­scape and com­mu­nic­ate more effi­ciently, espe­cially dur­ing times of change.

Learn more: The Silent Switch: How Algorithms Have Changed

Virality Through Real-Time Testing

A dom­in­at­ing fea­ture of today’s social media algorithms is real-time testing. 

If you pub­lish any­thing, the algorithm will use its data to test your con­tent on a small stat­ist­ic­al sub­set of users. If their reac­tions are favour­able, the algorithm will show your pub­lished con­tent to a slightly lar­ger sub­set — and then test again. And so on.

Suppose your pub­lished con­tent has vir­al poten­tial, and your track record as a pub­lish­er has gran­ted you enough plat­form author­ity to sur­pass crit­ic­al mass. In that case, your con­tent will spread like rings on water through­out more extens­ive sub­sets of users. 

This test­ing algorithm isn’t as math­em­at­ic­ally com­plex as one might think from a pro­gram­ming stand­point. The most robust approach to increas­ing vir­al­ity is redu­cing cycle times.

YouTube’s algorithm argu­ably does well in cycle times, but the real star of online vir­al­ity is the Chinese plat­form TikTok.

Read also: What You Need To Know About TikTok’s Algorithm

And this is where it gets pitch black. The com­plex­ity and gate­keep­ing prowess of today’s social media algorithms don’t primar­ily stem from their cre­at­ive use of big data and high-end arti­fi­cial intel­li­gence but from the blunt use of syn­thet­ic filters.

The social media algorithms could be much more com­plex using machine learn­ing, nat­ur­al lan­guage pro­cessing, and arti­fi­cial intel­li­gence com­bined with human psy­cho­logy neur­al net­work mod­els. Especially if we allow these pro­to­cols to be indi­vidu­al across ser­vices, we will enable them to lie to us just a bit. 

But, no. Instead, the social media algorithms of today are sur­pris­ingly straight­for­ward and based on real-time iter­at­ive test­ing. Today, vir­al­ity is con­trolled mainly via the use of added filters.

Underestimating the Effects of Filters

The algorithmic com­plex­ity is primar­ily derived from humans manu­ally adding fil­ters to algorithms in their con­trol. These fil­ters are tested on smal­ler sub­sets before rolling out on lar­ger scales.

Most of us have heard cre­at­ors on Instagram, TikTok, and some­times YouTube com­plain about being “shad­ow­banned” when their reach sud­denly dwindles from one day to the next — for no appar­ent reas­on. Sometimes, this might be due to changes to the mas­ter algorithm, but most cre­at­ors are prob­ably affected by newly added filters.

Please make no mis­take about it: Filters are power­ful. No mat­ter how well a piece of con­tent would nego­ti­ate the mas­ter algorithm — if a piece of con­tent gets stuck in a fil­ter, it’s going nowhere. And these fil­ters aren’t the out­put of some ultra-smart algorithm; humans add them with cor­por­ate or ideo­lo­gic­al agendas.

“There is no inform­a­tion over­load, only fil­ter fail­ure.”
— Clay Shirky

TikTok serves as one of the darkest examples of algorithmic abuse. 

Leaked intern­al doc­u­ments revealed how TikTok added fil­ters to lim­it con­tent by people clas­si­fied as non-attract­ive or poor. 

And yes, this is where the dark­ness comes into full effect — when human agen­das get added into the algorithmic mix.

“One doc­u­ment goes so far as to instruct mod­er­at­ors to scan uploads for cracked walls and “dis­rep­ut­able dec­or­a­tions” in users’ own homes — then to effect­ively pun­ish these poorer TikTok users by arti­fi­cially nar­row­ing their audi­ences.”
Source: The Intercept ³Biddle, S., Paulo Victor Ribeiro, & Dias, T. (2020, March 16). TikTok Told Moderators to Suppress Posts by “Ugly” People and the Poor to Attract New Users. The Intercept. … Continue read­ing

The grim irony is that adding fil­ters is rel­at­ively straight­for­ward from a pro­gram­ming perspective. 

We often con­sider algorithms advanced black boxes that oper­ate almost above human com­pre­hen­sion. However, with reas­on­ably exact algorithms, it is arti­fi­cial fil­ters we need to watch out for and consider.

Enter: The Electronic Age

Human cul­ture is often described based on our access to pro­duc­tion tech­no­lo­gies (e.g., the Stone Age, the Bronze Age, and the Iron Age).

According to Marshall McLuhan and the Toronto School of Communication Theory, a bet­ter ana­lys­is would be to view soci­et­al devel­op­ment based on the prom­in­ence of emer­ging com­mu­nic­a­tions technologies.

McLuhan sug­gests divid­ing human civil­isa­tion into four epochs:

“The Gutenberg Galaxy is a land­mark book that intro­duced the concept of the glob­al vil­lage and estab­lished Marshall McLuhan as the ori­gin­al ‘media guru’, with more than 200,000 cop­ies in print.”
Source: Modern Language Review ⁴McLuhan, M. (1963). The Gutenberg galaxy: the mak­ing of typo­graph­ic man. Modern Language Review, 58, 542. https://​doi​.org/​1​0​.​2​3​0​7​/​3​7​1​9​923

As a PR pro­fes­sion­al and lin­guist, I sub­scribe to the concept of the Electronic Age. I firmly believe soci­ety is unlikely to revert to the Gutenberg Galaxy.

Read also: The Electronic Age and the End of the Gutenberg Galaxy

The Power of Perception Management

This is the truth about how social media algorithms are con­trolling our lives:

No one makes decisions based on actu­al real­ity; we all make decisions based on our lim­ited men­tal maps (i.e. sens­ory “bits and pieces” stitched togeth­er of that real­ity). Hence, if you con­trol some parts of people’s per­cep­tions, you indir­ectly con­trol what people do, say, or even think. ⁵Lippmann, Walter. 1960. Public Opinion (1922). New York: Macmillan.

Social media algorithms influ­ence our men­tal maps of real­ity and, by exten­sion, our atti­tudes and behaviours.

What would hap­pen if Google and Facebook filtered away a spe­cif­ic day? Everything that refers to that day wouldn’t pass any iter­at­ive tests any­more. Any con­tent from that day would be shad­ow­banned. And search engine res­ults pages would deflect any­thing related to that par­tic­u­lar day.

“Since we can­not change real­ity, let us change the eyes which see real­ity.”
— Nikos Kazantzakis

To para­phrase a pop­u­lar TikTok meme, “How would you know?”

Learn more: Walter Lippmann: Public Opinion and Perception Management

The First Rule of Social Media Algorithms

Social net­works don’t want us talk­ing and ask­ing ques­tions about their algorithms des­pite being at the core of their busi­nesses. Because 1) they need to keep them secret, 2) they are blunter than we might think, 3) their com­plex­ity is mani­fes­ted primar­ily by arti­fi­cial fil­ters, and 4) they don’t want to dir­ect our atten­tion at how much gate­keep­ing power they yield.

And both journ­al­ists and legis­lat­ors aren’t exactly hard at expos­ing these appar­ent demo­crat­ic weak­nesses; journ­al­ists want their lost gate­keep­ing power back and legis­lat­ors because they see ideo­lo­gic­al oppor­tun­it­ies to gain con­trol over these filters.

Any PR pro­fes­sion­al knows that the news media has an agenda. And that we must man­age that agenda. Otherwise, it might spin out of control. 

A social media algorithm can be suc­cess­fully nego­ti­ated and some­times work for you or your organ­isa­tion. But an algorithm with its fil­ters will nev­er be your friend. 

Social net­works are “good” in the same way the news media is “object­ive”, or politi­cians are “altru­ist­ic”. As pub­lic rela­tions pro­fes­sion­als, we should act accord­ingly and man­age the social media algorithms — just as we man­age journ­al­ists and legislators.

Read also: Social Media: The Good, The Bad, The Ugly

PR Resource: More Social Media

PR Resource: The 7 Graphs of Algorithms