In 1896 automobiles or “motorcars”, looked funny and blacksmiths were not expecting that in a few years time, they would, for the most part, be out of a job, replaced by car workshops. If at that time everyone had a horse, now everyone has a car, and it is rare to know someone who actually owns a horse. At that time, some early adopters “simply gave up and used a horse” (Ridiculous U.K. Traffic Laws of Yore).
Many people have now tried AI assistants, some use it regularly, but many just gave up. Siri did not live to our expectations, and now we have Alexa or Google assistant that are slightly better, but we are still really far from a Jarvis like assistant even if every geek still dreams of it.
As once automobiles were, AI assistants are in their infancy, but things are changing quickly and it will be way quicker than what has happened with cars. First of all, AI assistants do not need a Henri Ford, as we do not have mass production issues, everybody already has a phone, the assistant is just a software update away. And people will have it everywhere, it will be integrated in their house, in their cars, in their headphones, actually not, it is already there. It is just not that common place yet.
What will happen? Assistants will be ubiquitous because they will be helpful and they will help us save money and time. It is not science fiction, it around the corner and 2018 is a pivotal year we have seen what Google is able to do:
The Assistants will change the conversation, and will make a strategic impact across sectors when they are able to understand on the same level as humans - or even better.
In the video, we see that Google Duplex – the most advanced assistant that has been publicly displayed – was able to understand not only the conversation, but that its interlocutor was not following the conversation, therefore the assistant decided to rephrase the question in order to achieve its goal make sure to have a table. Probably soon, hairdresser and restaurants will have their assistants speaking to assistants of customers. But this is just a stepping stone. To be helpful, AI assistant will need to know us: understand our preferences, our habits, our tastes, our values, what compromise we are ready to do, etc.
Assistants are not an utopia anymore. In a way their trajectory is similar to self driving cars. We hear about them, we did not see many of them, but they are around the corner. In 2017, 16% of all household in the United States had smart speakers, this number is expected to rise to 55% by 2022.
Google announced during the Made by Google event in October 2018, that Google Home queries were up 400% in the last year. It is predicted that 50% of search queries will be made by voice by 2020.
Last but not least, voice shopping last year represented $2 billions and it is expected to rise to to $40 billions by 2022. 94% of all devices currently in use are products from Google and Amazon. 82% of people owning an Amazon Echo have an Amazon prime subscription. This concentration of assistants in the hands of a few platforms is a problem for brands, retailers and other businesses.
In order to work an assistant needs a lot of data, it needs to know the preference of the user it assists, but also the preferences of the other users in the platform. When users tend to behave in the same way, we can assume that if a few users from that group like a movie or a product it should appeal to the rest of the users of that group too. This is called collaborative filtering and it is the backbone of classical recommender systems.
But assistants will collect prices and characteristics of goods and services, compare them, match them with users’ preferences. Because all of this data will be stored by the same platform, the more users use the service, the better it will be. Smaller assistants, with less people using them will not be able to compete as they will be less relevant. We have already seen this with search engines, before google we had many options, some meta search engine were even indexing results from other ones, but now we have Google, Bing and maybe DuckDuckGo, but everybody googles things… Google is the defacto search engine. The same thing will happen with assistants. We will see (and there are only a handful of platform already) a consolidation in this field with a few almighty platforms.
With the rise of assistants, marketeers will discover that users are not faithful. They think they are but, in many cases, it is a false assumption based on the fact that consumer buy again and again from the same brand, therefore, they are faithful to the brand. In fact, we, as consumer, are lazy. It is not worth our time to compare the composition of toothpastes and prices, and to do that again and again everytime we need to buy toothpaste. This is also holds true for bigger objects that we buy less often.
Maybe Volkswagen is running in the family, The grandfather was buying Volkswagen, the father was buying Volkswagen, therefore the son will buy a Volkswagen, because the family always buy Volkswagen, until? Someone or something shows the son another car, that has something new, something different, let’s say a Tesla, and comes with compelling arguments. Recommends articles and comparisons as well as detailed explanation on cost savings. That’s how a good salesman would convert the son, that is how an AI will convert him too.
It is also true for insurances. We do not buy insurances that often, when we do we kind of compare, but then, we just pay. Do we really spend our time comparing? Then we would have to do all the administrative tasks in a timely manner in order to change to another insurance.We basically give up, before even starting the process, because it is overwhelming to make those decisions all the time. AI will not be overwhelmed, and it will compare prices all the time without tiring, and when it is time for us to have a better coverage for a lesser price it will do all the administrative work for us. This is when marketeers, who believed consumer were faithfull will discover that they actually don’t care much about their brand.
In fact, when consumers start to trust the upcoming AI platforms, they will let them buy products on their behalf and the AI will look at the best interest of the consumer and not at the best interest of the brand. Why? Because those platforms will need the users to trust the AI to retain them. But let’s start by defining what an assistant actually is and how it will impact our lives.
What’s in an assistant?
Here is the definition of assistant that is of interest from the Oxford English Dictionary:
†1. One who is present, a bystander; one who takes part in an assembly. Usually in plural. Obsolete.
2. One who gives help to a person, or aids in the execution of a purpose; a helper, an auxiliary; a promoter; also, a means of help, an aid.
The second definition is what we understand by an assistant, an aid in the execution of a purpose, but I like the obsolete definition: a bystander, one who is present. This seems to be more and more true for digital assistants, they are an aid, but they are always present, now they are even always listening, they take part in human assemblies.
In our digital world, an assistant, to be truly an assistant, has to be much more than a search engine with voice enabled, it has to be able to anticipate our needs and prevent us from forgetting things or missing events. It needs to take in consideration our means and our habits but also our tastes and preferences. It means that an assistant to be an assistant needs to know us, it needs to have access to our behaviours, to our money but it also needs to be able to act on our behalf.
To be performant, an AI Assistant is a an advanced form of contextualized recommender system. In this case, contextualisation is not only what we are doing at the current moment, but much broader, since it will need to understand time, money, social relations, career; all the things we care about in our lives
Our lives are ruled by time, and we have the faculty to reflect on what we have done, to use our history as experience in order to behave or do things differently at the present time or in the the future. An AI assistant will have to understand that properly to be really useful.
Why do we set an alarm clock? Because in the evening we know where and when we need to be in the morning and we estimate based on our knowledge when do we have to leave home. We estimate the time we usually take to have breakfast, to wash our teeth and take a shower, so we retroplan our time to wake up based on this information.
An AI could do better and set this time for us, it would have to learn our habits, but it could then take in consideration the time we need to sleep, to suggest that now it’s time for us to go to bed. If during the night something happens, maybe it starts snowing and roads are congested, or anticipated to be congested, it could change our alarm and inform us in the morning of the problem.
Of course, the temporal context does not stop there, the AI needs to understand the time of the year we live in, maybe it should reserve some time for christmas shopping (and suggest some presents). The AI will know that if we travel somewhere in a few weeks, we might need a jacket, but it will also order for us an Uber and the necessary train tickets.
An assistant, will also manage our money, it will buy things for us to refill the fridge, order toothpaste when it knows that we are running low, it will buy train and airplane tickets. But it will also help up us save money. This is what will be most disruptive for brands and I will get back to this. But in short, it will recommend products, or even buy them directly. It will do this by comparing prices and match it with our preferences, maybe a new bio toothpaste appeared on the market and is cheaper than the rest of the established brands. The AI will compare characteristics and propose it to us.
It could also manage financial services, switch our bank accounts to another that is cheaper (thank you GDPR and data portability) and do all the paperwork for us, inform the company we are work for, and check that we are indeed receiving our salary etc. It will compare insurances and select the best one for us. It could help us save money by rounding up all our purchases and putting all that money in a savings account. There are already online banks that propose these services, but they will probably be incorporated in the AI platform directly.
Human relations and digital workplace
The AI will help us with relations. Thus we will never forget the birthday of someone that counts for us. The AI will even be able to suggest the best present, as of course it knows us, therefore me but also the person who has a birthday.
Human ressources start to use AI for a range of tasks from business automation to candidate selection or even reduce administrative load. On the other hand, candidate are starting to learn how to beat AI to get a job… This might all change and get somewhat simpler. Our assistants will know us, know our skills, know what we are capable to deliver and which environments will fit us best. What if the assistant took care of our career? It would suggest courses to follow online that fit our needs, suggest meetups and conferences where we could learn something for our career, or even meet someone who could advance our career.
Then, on a nice Monday morning, the assistant would tell us that we have resigned from our current job, and that we are starting a job at another company. Of course, that job is better paid and with more perspective than the one at our previous company.
The assistant will work in the background to apply, send our resume, answer questions on our behalf, and finally when it get’s the contract, it would resign on our behalf at our current company. The assistant would have “spoken” to the company assistant and no human would have been involved in the process.
The assistant could have even anticipated that it is time to leave our company, as it forecast bad times ahead for it. Indeed, our assistant follows the news, the new laws that are enforced, potential lawsuits that would affect our company, the stock market, and because we are insiders, it could understand that something is wrong with the business, customer are less happy, they churn, and therefore, it it time to leave.
All these scenarios will work if users learn to trust those assistants. Trust will be build in a number of different ways and we will see that the way AI assistants, or let’s call them AI platforms, building trust will actually be detrimental to brands and how they were used to selling their products. Indeed, when people trust in these platforms, it will revolutionized the way we consume.
We will integrate AI platforms slowly in our lives and everytime the experience will be successful, we will trust them more, thus we will delegate more. It will be a virtuous circle. In the other hand, every bad experience will make us think of the assistant as a toy and we will not trust it. Siri is a good example of that. People found it cool, but then, that was it. It now lags behind Google Home and Alexa.
In order to build trust, AI platforms will have to be relevant, to satisfy users and to match their needs. As we have already mentioned, brands will discover that consumers are not faithful, they are and might still be for brands that are associated with social status or that gives a sense of belonging to a group but this factors will be integrated in the recommendation systems. We could even argue that fashion, will not be dictated by brands (or designers) anymore, but by trends and recommendations…
We can see that with the press influencing buyers through pictures of stars caring some fancy bag or wearing cool glasses. We think that there is a designer behind the style of star, but it could as well be product placement or advertisement. The recommender system will find the correct object in the correct picture that will appeal to the correct user, and it might even sell information to brands telling them that if they want to influence this segment of the population, they need to place a product with this actor or that singer.
As users, we already search for articles describing the current trends in order to guide our choice of hair or nail color, to find what is fashionable and trendy. Assistants will be very good at it as they already have all this information. Amazon knows what color is the most bought in 2018, and it knows what people are buying. Facebook and its subsidiaries like Instagram, Messenger, or WhatsApp know what we are wearing, what we are eating as we still have an enormous appetite for taking pictures of our food and posting it online. They also know what we are talking about.
The fact that data gathering at this level is creepy is beside the point. Users are happy to share data for a little gain, while at the same time stating that they are against sharing private data. In this context, assistants will earn our trust, step by step. We can already trust Google Assistant when we get a notification telling us that the last train leaves in 10 minutes. We had to entrust some bits of informations such as our home address, but then the assistant is able to infer that we are not at home, that we usually use the train as transportation, that we might not be able to get home, therefore that we might need to know that information. We have shared our privacy, but the information is valuable and therefore we are satisfied by the service. When Google manages to match our needs of transportation and prevent us from having a problem It makes us happy, we are impressed and we start trusting the assistant. As new capabilities are built in assistants, and with their abilities to successfully deliver new experiences our trust in them will only increase.
Brands and AI
We have seen that assistants will have to be relevant. What does this means for brands? The direct relation brands have with customers will be replaced by a direct relation with assistants. In turn, assistants will have a direct relation with customers. AI platforms will be like a best friend who recommends a product. We value this kind of recommendation and we will value the recommendation of the AI platform. We will prefer that to a plain advertisement or a two for one bundle.
The danger for brands is that these AI platforms will learn the relevance of a product and will increase its weight (upvote it) and downgrade the product that has been replaced. It will be based on the behaviour of the user and a successful recommendation by the assistant. If the assistants recommends a brand instead of another one, and the user accepts that recommendation, this recommended product will be seen as a better option for that type of user. Therefore the collaborative filtering algorithm (recommendation based on the choices of similar users) will start to recommend that product to other people, and one brand will replace another. We will witness massive and sudden change in revenues for brands, and they will have no control on what happens with recommendation as brands are out of the equation.
Not only assistants need to be relevant, they also need to match the needs of the user, and for that, they cannot betray the user/customer. If they do, the user will lose faith in the assistant and the whole system will crumble. To match the needs of users is a source of conflict of interest for the assistant as they need to keep the trust of the user and to satisfy their needs, but they could be tempted to make money by selling placement (as it is now done on search engines and other comparator platforms).
Indeed a pure commercial approach for assistants would be to copy the current model and to ask brands to pay for visibility, but it will not be long before the user will feel that the game is rigged, especially when it is so important with voice interfaces to be the first product in the result lists. Especially because the length of this list will, most likely, be one. Therefore assistants cannot use the current advertisement business model. In the best scenario, the assistant could be opened about advertisement and answer to a query by saying that there is this product that is currently being advertised and another one that people usually choose. This solution is not satisfactory as the brand is not seen in a very convincing light. Therefore, assistants will probably chose another business model.
AI Platform Business Model
AI Platform to work will need a huge amount of data not only on users, but on brands, on products and services, on the world. Thus Google, Amazon, Alibaba and other big tech companies have a very good head start in this race. AI Platforms will satisfy user needs. As we have seen, this will be built in. It is already difficult for brands to measure customer satisfaction, they do survey or try to infer satisfaction from statistics but this will get even more complicated for them as they will not have access to the customer anymore.
But AI Platforms will know everything about customer satisfaction. They will know all ratings, all reviews, all discussions, but also they will have detail stats on different funnels and customer behaviours. AI Platforms will know when a customer did a search, then went to see the product in a store (at least they will know that the user went to a store that corresponds to the product searched, therefore it could be inferred that the user wanted to see the product), then searched some reviews, then compared prices. The Platforms will learn from these behaviours, they will also provide on site tracking systems in order to help stores better understand customer behaviours, but in the end this will only enrich their contextualized recommender system.
If now brands have multiple channels to reach out to customer, this will boil down to one or two main AI Platforms that dominate. It will be even more difficult for brands to do market studies. Indeed the consumer behaviours, decision making will be filtered or dictated by some kind of algorithm. As we will see, it will also be difficult for brands to understand their competitor pricing as well as their own pricing.
Nowadays it is difficult to target customers. Brands have to identify billboards, newspapers, TV shows, websites, etc. and try to mix these potential “location” with data that was gathered on different advertising platforms, retargeting platforms and other analytics data. These targeting solutions are not perfect and rely on experience and experiments.
AI platforms will have to serve the interests of users – if not they will not gain their trust and they will not be relevant in the coming economy. These platforms will also position themselves as a service platforms, in the manner of credit cards. Clients have to pay a fee for their credit cards, whether it is included with their bank account fee, or as an additional fee. Stores have to pay a cut to the credit card platform for the service and they also rent devices to read credit cards. Credit cards is a service that stores have to have and there are only a few credit cards that are globally accepted. Mostly we see Mastercard and Visa. There are other credit card companies, but they are marginal and answer to specific needs. It is true that we see a new kind of bank appearing on the market, such as Revolut or Monzo that offer alternatives to the established Credit Card players, but in the end their cards are MasterCards or Visa.
Credit cards are in the middle and earn money from the customer and from seller, they are in a very good position and push for their system to be easier and more convenient to use than cash. If customers are using cash, credit card companies have no cut. Therefore, they have an interest in removing cash. Thus they provide bonuses, pay back, insurances, and show the safety advantages not to have a cash on you.
As a side not, this is why Bitcoin is so popular, it position itself a true digital cash. When there is a money exchange with Bitcoins, the money, as with tangible cash, goes from one wallet to the other, with no third party involved in the transaction. Bitcoin allows exchange of money without middle man, therefore it makes Bitcoin a real money, a money that, as with real cash that cannot be traced (unless a user is not careful and leaks is id and IP address.). This is why government do not really like it as that money escape taxes as when there is a cash transaction in real life. It is also why the typical financial institution are feeling threatened by real digital currency as there is no money to be made.
Back to our subject, platforms in general are positioning themselves as middlemen. This position allow them to disrupt whole industries. We think of AirBnB or Uber which are respectively the biggest hotel chain in the world and the biggest taxi company but that do not own any room or any taxi, but are able to generate huge revenue by organising reservation and taking a cut on transaction. In a way this is the same model as credit card companies.
AI Platforms will also position themselves as the middleman, and as the consumer’s best friend will know everything about the him or her, but also the stores or services. They will have knowledge on markets and consumers. It is already in 1597 that Francis Bacon wrote in Meditationes Sacrae and Human Philosophy : Knowledge is Power. Therefore this privileged position will let AI Platforms sell this information to brands.
AI platforms have, or will have, an extended knowledge on what customers are doing and who they are. They will as credit card companies sell that valuable information. But they will not stop there. As they grow these AI platforms will offer more and more services such as banking or leasing. We can already see this evolution with Uber providing credit cards and car leasing while this is not its core business. Google and Apple are moving in the payment sector. It is only logical. When Uber cannot expand because drivers do not have cars, they have to help them get one to expand their business, if they do not have a bank account because it is difficult to get one in their country, Uber helps them. These platforms do that because they do not want to be slowed in their growth because of legacy systems and administration that cannot scale as needed. There is no reason for AI platforms to behave differently.
Source of information for AI assistants
For marketeer, and more generally for Brands these AI platform will be a source of information. That information will not be free, but by paying, they will get very detailed segmentation and price range to target and sell their products.
Brands will be able to know who their customer really are, this information will be representative of reality, especially when assistants are ubiquitous. They will not need, as they are doing now, to invest in fidelity programs, bonuses or point systems, or at least not in the same manner, or for the same reasons. Brands will have to reevaluate how they keep a relation with their customer and if they need that relation at all. They will have to be sure that at the moment when a potential client wants to change his car, their vehicule gets recommended. The easiest way to do that will be to to have a relation, with the AI platform and not necessarily with the customer directly. Especially if the recommender system recommends the car, and buy it from the brand in order to lease it to the customer.
Despite the fact that some studies show that users could even like personalized ads, brands will need to reconsider their investment in ads, since everybody hates ads. People in the ad business try to convince themselves that people hate only bad ads. But in the end, why do people install ad blockers? Probably because they do not like ads, or at least they do not like pay-and-spray tactics.
But for now digital ads work, and spendings are increasing. Although Alphabet is diversifying its sources of revenue, in 2018 Q2 86% of Alphabet’s revenue is still ads. And year to year the revenue increase by 25% for Alphabet. Facebook saw an increase of 42% in ad revenue thanks to its mobile ads. In 2017 digital advertisement as surpassed TV ads for the first time and digital advertisement is seen as key growth in the business landscape. In 2017 33% of ad revenue, was going in to Alphabet pocket while Facebook was taking 16% and Alibaba 8%. As digital ads market share is increasing, the revenue is also concentrating in a few hands. 57% of the digital ad revenue was shared among 3 companies, and the following ones are Baidu, Tencent and Microsoft are unsurprisingly also tech companies.
But ads are programmed to disappear, at least in their current form. Indeed with voice enabled systems, there will be no real estate for ads, no room to insert ads, and who will want an assistant that interrupts its discourse to speak out loud an ad? Marketeers are currently learning that they need to personalize ads on advertisement networks such as Google’s, Facebook’s or Alibaba’s. But also to personalize the user experience at all levels. Personalization will help big companies to lock brands in as they will more and more rely on data that will be difficult to collect, while it will become easier and easier for the big tech companies behind ad networks, social networks, e-commerce and other platforms to integrate the data they collect and use it in assistants.
AI platforms will become a source of information for brands and services, who will have to pay the platform to know how much they should price their product to reach a certain segment of population. Businesses will have to find a way to target AI platforms, or to target their products for those platforms and not necessarily to target customers directly.
Brands will need to have reactive pricing. They will need to change their offering for the AI platforms to notice and consider the new pricing a better option for customers. In order to change their pricing, brands will have to buy information from the platforms. This will be a paid service, but in the end, it will, probably, be the first time ever that stores or services will really be able to know who their customer really are without relying on vague market study and approximate surveys. Brands will pay for the access to big data, which they will be able to drill down in order to really know what their market is. They will be able to create better segmentation, ideally have a per customer segmentation with a personalized offering for that customer. In the end, prices might be unique for a specific customer at a specific time.
AI assistants as a new kind of bidding platform
As we have seen it is most likely that AI platforms will not sell ads. Their business model will be centered around data and selling information. But it will not be limited to that. AI platforms will be able to create a new type of bidding platform based on alerts.
Current bidding platform, such as Google adwords, let companies fight over key words and positions. But since AI platforms will not push ads, people will not search with keywords. However, the AI platforms will know, as they are the ones who is starting to recommend one product instead of another. As such, they will be able to alert a brand that its product will not be recommended anymore for a specific segment. This alert will allow the brand to create a counter offer. Resulting in a very competitive market place that will benefit consumers.
This means that brands will have to be able to automate that process in order to be able to react to alerts send by AI platforms. It seems that it will not be possible to do that manually at scale. This will enable brands to undercut their competitor, but for a price. We could argue that brands are already spending huge amount of money on bidding platforms but the nature of that bid will change, and brands have to prepare for it. It means that they have to be very flexible with pricing.
It will be interesting to see what brands develop to automatically react to offers from their competitor. As brands will want to keep a competitive edge, they will probably develop an algorithm that will be fed by they information bought from the AI platform and we might see automated systems fight against each other. It might become a lucrative place for traders to invest.
Role of Brands
In this changing new AI economy, brands will have to be aware of these upcoming changes, to look out for tweaks of the AI algorithm, and they will have to study them because it will be the primary way of interacting with consumers. One of the new digital job that appeared was Search Engine Optimization specialists. People who were trying to figure out how the indexation and ranking algorithm of search engine works. It is a job that requires to study competition and do a lot of experimentation. The coming AI platform will also create such jobs and brands will have to be ready to learn and experiment from the onset of these platforms to make sure they stay relevant. The start will be slow, but those who will understand how to position themselves will be in a much better position.
Stores and services will also need to reevaluate their physical location as most likely it will not be necessary anymore to keep a show, workshop or an office. Will drivers bring their car to the car workshop? Probably not, as the car will drive itself there. Hairdresser will come to the office or at home. Food will be waiting for us at home. At some point it seems physical stores will be useless. Unless, businesses can reinvent themselves and deliver an experience that AI platform cannot provide: empathy.
Brands should find a way to keep a relation with customers, they need to use another skills that AI is not good at: imagination and (true) creativity. In order to forge new experiences and new ways to keep a relation with each customer. Brands have to bring the AI on a terrain where it does not have the advantage. AI is good at data and numbers, but not at anything that makes us human.
At the same time theses AI platforms will create many opportunities, new jobs will be created for those who want to understand how AI platforms work. Probably this will spread into different type of jobs as we can see now with Search Engine Optimization, Search Marketing Optimization, Social Media Marketing, Pay Per Click specialists, etc. As for any new area, there will be no training and people will learn by experimenting and experience will have to be demonstrable.
But there are also opportunities in the development of “trading algorithm” and in finding new ways to benchmark the the ever changing pricing of competitors in order to align and beat the prices of competitor. But brands, or any startup that wants to enter this new field, have opportunities in directly and continually studying the recommendation algorithm of AI platforms in order to be ahead of the platform and competitor, or, if it is a startup entering the field, to sell that information as a service.
Of course we could study these platforms manually but the most efficient way would probably to create some kind of half Generative Adversarial Network (GAN) and to maintain control profiles. GAN is an unsupervised form of artificial intelligence where two neural networks confronts each other in a zero sum game. The sum is always zero, therefore if one has the 1 as a result, the other loses with -1. The idea would be to create a neural network that would generate user profiles and behaviours in order to create realistic assistant traffic that would get accurate recommendations from the AI platform. The generative network would win, when a product of the brand gets recommended.These profiles would be constantly updating and brands would know how the AI platform recommends products and therefore could anticipate changes. The generative network would save “wining” profiles and use them to control what the AI platforms recommends to these once winning profiles, but would continue to create new profiles to monitor what is happening with time.
The task is not easy, but it could be really worthwhile. Brands have to understand what is coming and prepare for it. They have to understand that they will have to change their way of advertising, and redistribute their spendings in new areas such as specialist who will take time to understand the upcoming algorithms, and or in research to find new ways of advertising and selling their products if they do not want to be left behind. The half GAN approach is one idea, it is just an idea, but it is doable it is as much an opportunities for start up and for brand to be successful in the new area of “high speed trading”, I meant shopping.
Yunhe Pan, “Heading toward Artificial Intelligence 2.0”, Engineering, Volume 2, Issue 4, 2016, pp. 409-413
Sheridan, T.B., 2018. “21 Speculations on Future Relations Between Humans and Automation”. Automation and human performance: Theory and applications, p.272.
Spyros Kokolakis, “Privacy attitudes and privacy behaviour: A review of current research on the privacy paradox phenomenon”. Computers & Security, Volume 64, 2017, Pages 122-134.