bot-or-not-1.png

InternetMay 6, 2024

Here’s what happened when I replaced my personality with the Meta AI chatbot for a day

bot-or-not-1.png

Keeping up with online communication can be exhausting, so Fran Barclay enlisted the help of Meta’s new ‘intelligent assistant’ to respond to all her messages. Could her mates tell the difference?

For centuries, technology has ruled the ways in which we communicate. From the dawn of written language, to the printing press and the internet, our interpersonal relationships have evolved alongside technological progress.

In the latest instalment of this ancient symbiosis, Meta’s new chatbot, Meta AI, has arrived. Heralded as an “intelligent assistant”, the tech giant claims that its artificial intelligence will “expand your knowledge and capabilities”. 

As someone generally incapable of keeping up with social media messages, I put this assertion to the test. Could the bot effectively replace me in my digital friendships? For most of us, keeping up with loved ones is a key motivator for using social media. Can Meta AI help us to maintain these valuable connections by taking the effort out of crafting an “intelligent” response?

As you may have noticed, the bot is now embedded in WhatsApp, Instagram and Facebook Messenger. You can chat with Meta AI directly, like your childhood imaginary friend – only this one feeds your secrets to one of the most powerful companies on the planet. Alternatively, you can tag it into group chats and private messages like that Grecian oracle you’ve always wanted to consult. 

In a bid to save my relationships from a slow and painful death by unread messages, I asked the chatbot to write my responses for a day. Of course, this would only prove a viable solution if my pals couldn’t tell the difference. As ethically as possible without their explicit consent to this process, I set about copying and pasting messages from friends into my private chat with the algorithm, where I requested an appropriate response. Within a few seconds, the chatbot would generate perfectly personable answers. What could possibly go wrong?

‘Creepy as hell’

Unlike in the 2013 AI movie Her, large language models are not trained on our personalities – they’re trained on the internet. That means anyone who writes with a degree of personal style, slang, or even trademark typos is going to sound unusually generic when Meta AI speaks on their behalf. 

To be (human), or not to be: that is the question

Faced with a message from a former flatmate, I turned to AI to keep the conversation flowing with a suitable dose of enthusiasm and rapport.

Nothing like a bike ride to bond with friends

Fantastic news indeed! And even better was the speed with which I could show my delight at her goal-crushing antics, with a brief ad-lib about my own week. 

Clearly this bot was invested in my friendship if it said we should meet up IRL. Still, there was something deeply unpleasant about outsourcing communication with someone I used to see on a daily basis. Yes, by replying I was showing my desire to stay connected, but it came at the cost of sounding strangely like Mark Zuckerberg – or as this friend put it, “creepy as hell”. 

Fortunately, Meta has some sage advice for customising content to your style and friendship:

  1. Know your friend’s sense of humour: If your friend appreciates sarcasm or playful teasing, use that to your advantage. If they’re more serious, tone down the humour.
  2. Reflect their language and tone: Mirror the level of enthusiasm, formality, and language your friend used in their initial message.
  3. Reference shared experiences or inside jokes: This helps create a sense of connection and makes your response feel more personal.
  4. Be authentic and genuine: Use your own voice and avoid using overly generic or cheesy phrases.
  5. Keep it concise: Don’t overdo it with a long, elaborate response. Keep it brief and to the point.
  6. Pay attention to your friend’s preferences: If they mentioned specific interests or goals, show interest and ask follow-up questions.
  7. Use emojis wisely: Emojis can help convey tone and friendliness, but use them sparingly to avoid coming across as overly enthusiastic or insincere.

As the chatbot said, the goal of any relationship is to “be yourself” and show “genuine interest.” Words to live by in this age of disillusion and ennui. 

‘I’m gonna unfriend you’

Now that I was better equipped to work with this technology in a more collaborative manner, I tried these tips on a former colleague, Dan. 

Dan had previously advised that he finds Meta AI “boring” and would not tolerate its use in our relationship. Meta’s AI policy asserts that the bot should not be used to “deceive or mislead others”, so I turned to the algorithm to craft a response to this ultimatum:

Nothing more old-fashioned than instantaneous communication online.

Either this was just the bot’s attempt at “playful teasing”, or Meta AI had breached its own ground rules. I did my best to keep the deceit “concise” and match Dan’s language, but true to his word, he blocked me for the rest of the day. 

Still, there was something endearing about the concern with which the algorithm tried to counsel my relationship:

 “I hope your conversation with Dan goes well and that you’re able to stay in touch in a way that feels authentic and meaningful to both of you,” it said.

If that all sounds incredibly two-faced, that’s because it’s a machine. As Allyn Robins, AI lead at the Brainbox Institute, explained to me, Meta AI “doesn’t ‘understand’ things” and nor does it have a moral compass.

“Don’t expect consistency, or for it to be particularly useful for anything that requires specific knowledge or authenticity,” Robins warned. 

Missed opportunities

Authenticity is a nebulous goal in digital communication. What does it mean to be “real” on a social media platform known to manipulate our thinking patterns and behaviours? Most of us don’t have a clear idea of who we are in the eyes of our beholders, but we do exert some agency in how we express ourselves – even if it’s through a well-placed emoji or a meme. 

By the end of this experiment, I was undeniably anxious about the harm I could have caused to my relationships. At the same time, I was encouraged by the ways in which the algorithm seemed to reinforce the importance of human connections. When I asked the chatbot to articulate the risks of what I had been doing, it listed “missed opportunities” for “meaningful moments of connection and deeper conversations with your loved ones”. 

Arguably, that means getting off the internet and trying to see said loved ones in person. But for those of us who live too far from home or are fiscally obliged to make a living, digital communication will continue to play a key role in maintaining our relationships.

According to Meta, its artificial intelligence is meant to “augment your relationships, not replace the authenticity and emotional depth that comes from direct, human interaction”. It seems that being ‘authentically’ yourself, with all your problematic punctuation, is one function that AI can’t yet replicate.

‘Media is under threat. Help save The Spinoff with an ongoing commitment to support our work.’
Duncan Greive
— Founder
Keep going!
a supermarket checkout where a woman in grey smiles as she scants her card - bunt there is a red box behind her and a threatening vibe, like she's being watched
Scanning a loyalty card means discounts, but it also means surveillance. (Image: Getty; additional design: Tina Tiller)

SocietyMarch 25, 2024

Supermarkets know everything from your gender to licence plate number. Do you care?

a supermarket checkout where a woman in grey smiles as she scants her card - bunt there is a red box behind her and a threatening vibe, like she's being watched
Scanning a loyalty card means discounts, but it also means surveillance. (Image: Getty; additional design: Tina Tiller)

CCTV cameras, self-checkout machines, data-gathering loyalty cards, facial recognition: supermarkets gather an obscene amount of information about customers. Can they be trusted with it? 

Every day for the last few weeks, J Frank*, an Auckland software developer, has checked the letterbox. He’s been waiting for mail from Woolworths containing their new Everyday Rewards cards, the supermarket chain’s loyalty programme.

Concerned about how much information the supermarket has about him, Frank has devised a way to get the bonuses of loyalty discounts without the big data tradeoff: have four or five different cards under fake names, which he takes turns switching out with his other flatmates, so no one person has the same buyer activity associated with them. Living near two Woolworths supermarkets, he goes to the supermarket chain a lot – but while they might know his address, he’s determined that he won’t have his individual data monetised by the chain. 

“My main concern is where this data is going, and who else is using it,” Frank says. It might seem that his shopping information is innocuous – a bunch of bananas here, a bottle of milk there – but he doesn’t like the idea that this information might linger for years on unseen databases, potentially being shared without oversight, beyond his control. Even if the information is secure, it’s the principle of the thing: he wonders what the supermarket chain gains from having individual information about their customers. 

But his qualms aren’t quite enough to stop him from wanting the discounts the Everyday Rewards card offers, so the fake names, and the somewhat reluctantly recruited flatmates, are his compromise. 

a supermarket entrances with everyday rewards bunting
Supermarket bunting welcomes the new Everyday Rewards card (image: The Spinoff)

Frank’s card shuffling might seem paranoid: most people probably have at least a few loyalty cards loitering in their wallets, filled with bookshop stamps or outdoor store discounts or fuel savings codes. By sheer virtue of frequency of visit and variety of products available, the supermarket card will probably have more information about you than a coffee shop or fuel station – but is it really anything to worry about? 

The discounts on a loyalty card are just a sweetener, says Gehan Gunasekara, co-founder of independent group the Privacy Foundation and a law professor who has worked with students researching loyalty schemes. What makes that scuffed bit of plastic really valuable, fake name or no, is your data. “They know everything you buy, every item, which store it’s bought from, how often you go to the supermarket, which brands you like,” Gunasekara says. “That’s really fine-grained data – it can tell the story of your diet, your preferences, your whole life.” 

Getting people to care about privacy can be an uphill battle, which Gunasekara is very familiar with: after several decades of business being increasingly digital, surrendering your personal information, voluntarily or thoughtlessly, has become so prevalent that the event itself is quotidian. When you click “accept all cookies” or tick a terms and conditions box without reading said terms and conditions, when you tag your location on an Instagram post or look up your local public transport timetable, you are placing another coin into the dragon’s hoard of digital information. 

No-one is immune: Vladimir Putin’s location has been deduced from the ads served to the phones of his immediate retainers; secret American military bases have shown up as hot spots of digital exercise data on Strava. 

At the supermarket, though, the exchange of personal information for food is more tangible. At the entrance, you might be greeted by a sign with small text alerting you to new cameras or a facial recognition trial. Look up from the bewilderingly expensive limes, and you’ll spot dozens of CCTV cameras on the ceiling, like little black blisters. Increasingly, cameras hover above the self checkout too. And if you want those member prices – which can be significantly cheaper than the normal ones – don’t forget you’re giving away some information when you swipe your phone or card against the scanner. 

Gehan, a brown skinned man with a slightly receding hairline and a a friendly smile grins in front of a brick wall
Gehan Gunasekara says loyalty cards are a way for supermarkets to get more data about their customers (image: supplied)

On one hand, maybe this is fine. After all, most data is not used to identify and target individuals; revealing to some faceless goon in an office that you, in particular, have a weakness for Monday kombucha. Instead, it’s embedded in automatic processes, used in aggregate. Some serve a direct benefit to consumers: if the data shows that more people are buying a brand of chocolate biscuits at one supermarket, then that store can order more to keep everyone in stock. Your buying patterns could also be used to recommend particular products to you if you’re logged into the website or the app. And while discounts might simply be a sweetener to encourage people to sign up to membership cards, they do offer genuine savings when many people are struggling to afford food. 

“We use non-identifiable and aggregated data gathered through the Clubcard programme to better understand our customers so we can place them at the centre of our decision making,” a Foodstuffs spokesperson says (Clubcard is New World’s customer card; Paknsave doesn’t have one). Data collected through the Clubcard programme isn’t sold to third parties, although it is shared with partner organisations like Flybuys. 

‘Like a thousand-piece jigsaw puzzle, each member is vital to the whole picture. Join today.’
Calum Henderson
— Production editor

Woolworths say that their card data is well protected, too, despite Everyday Rewards being dogged by privacy concerns, including concerns about the way the card can link your car’s licence plate to your shopping. (Update: Woolworths has clarified that license plate data is not linked to Everyday Rewards cards.) According to the Everyday Rewards privacy policy, customer data can be used “to understand shopping habits and likely preferences”, enabling direct marketing to the customer. It can also be shared with Google and Facebook for advertising and data matching, as well as analytics companies and market research providers (details aren’t provided on what providers these are, and what their privacy polices are). 

Of course, you’re only going to know this if you read the terms and conditions in full, instead of just ticking the box to say you’ve read them. 

a cctv camera on the ceiling of a supermarket
If you look up at the supermarket, you’ll spot dozens of CCTV cameras (Image: Shanti Mathias)

In person, customers have to opt-in to using the card. This can feel galling given non-member prices can be as much as 1.6x more than member prices for items like canned tomatoes. For people who shop online, there’s no choice at all: Woolworths’ online ordering systems are now completely integrated with Everyday Rewards, so that you can’t order online without being signed up. New World online shopping is also integrated with Clubcard. 

“We have to ask if consumers are getting value for the data they’re generating,” says Gunasekara of the Privacy Foundation. 

Gunasekara says that companies should be trying to minimise the data they gather from customers, and that the data that is collected should be reasonable. You can’t get online shopping delivered without giving the company your address, for instance – but do they really need to know your gender identity? 

supermarket shelves with a big everyday rewards sign saying 25% less on it
Reward card discounts are a sweetener – but your personal information is even more valuable (Image: Shanti Mathias)

Once data has been gathered, it needs to be securely stored, which can be a liability for companies, because if the data is valuable to them, it’s valuable to others. Recent data breaches in New Zealand and overseas have targeted DNA data, vaccination information and personal loan accounts, and those are just the ones that are high-profile enough to be reported on. (Both supermarket chains say data security is important to them. “We have cybersecurity measures, encryption, access controls as well as administrative measures such as robust policies and procedures for security, data and privacy,” a Woolworths spokesperson says.)

While data generated during your weekly grocery shop might seem less important than information about your health or finances – surely your personal craving for balsamic vinegar chips doesn’t mean much? – it’s powerful in aggregate. Dubbed “surveillance capitalism”, the ability to collect, store and use data at big volumes has been enabled by digital technology, and Facebook and Google are some of the world’s most profitable businesses because of how much information they have about what people are into. Supermarkets are only doing what all these other businesses did first.

Loyalty cards are only one piece of the bigger puzzle of surveillance. Neither supermarket chain confirmed how many CCTV cameras they have on average, but in Australia, that number is 62. The future of supermarkets likely has more cameras, not less: Foodstuffs is currently trialling a technology that allows facial recognition in stores for trespassed customers, and Woolworths is adding extra cameras to self checkouts to ensure items are scanned correctly.

a sign at a supermarket informing people that there is a camera trial at the self service checkout
A sign informs supermarket patrons a camera trial is going on – but when supermarkets are your main choice for food, can you meaningfully consent? (Image: Shanti Mathias)

Supermarkets aren’t the only companies that use CCTV: there are at least 10,000 publicly owned CCTV cameras in New Zealand and as many as 400,000 privately owned ones. But supermarkets are different because they provide daily, essential nourishment – something much harder to avoid than a targeted ad for a fancy face cream chasing you across the internet. “The kind of leverage these companies have to collect information is very one-sided,” Gunasekara says. You need food, and in many places one supermarket or the other is the only place to get it, so privacy concerns come second to necessary sustenance. 

Given how widespread surveillance is, Frank’s persistence in using fake names for his Everyday Rewards card seems almost quaint, one small effort at resisting a system that takes a torrent of customer information as a given. Even as someone exceptionally aware of privacy violations, he can’t resist the discounts from an Everyday Rewards card. Think of all the people who are truly struggling to afford groceries. Is a choice to give up some privacy in exchange for slightly cheaper food a choice at all? 

In the digital world, the data harvesting and surveillance we have been putting up with for decades is silent and seamless. At the supermarket, it’s much more in-your-face. There it is, in the black eyes of the CCTV cameras. There it is, in the weekly email of recommended products from New World, your face briefly blinking on a screen before being deleted if it’s not a match. There it is, when you’re swiping your rewards card at the self-checkout machine, handing over your personal data before you can take your food home. 

Could the increasingly surveilled supermarket be the catalyst that finally makes us care about our privacy?

*Name has been changed to protect – you guessed it – privacy. 

This article has been updated to clarify that Woolworths does not link license plate data to Everyday Rewards cards.