Customisation by Artificial Intelligence: from the communication to the product
Fast fashion 2.0 and product as a service (PaaS)
p. 363-371
Texte intégral
1“The rapid and widespread development of digital services has been at the heart of the digital changes that impact our lives”, stresses the recent Digital Services Act1. All countries embrace digitisation and important aspects of humans’ lives have moved online. The proportion of e-commerce has increased dramatically during the last few years, particularly during the coronavirus crisis, and customers who are now accustomed to the use of the networks, will not be easily fooled. They are becoming jaded and it is more and more difficult to catch their eye and grab their attention. An attention economy is emerging. How to channel the Internet users’ attention is the major issue. These Internet users fortunately (unfortunately? ) leave huge quantities of data when they use search engines, visit web sites, subscribe to an online service, use social networks or shop in market places, exchange messages, click “like”, accept cookies,… Data analysis delivers valuable information on the customer’s needs, tastes, dreams, plans… This information can be of great help for designing the form of communication used with them and for determining the best ways to attract their attention (Bruno JULLIEN, 2021).
2“With the evolution of the Internet that made tracking of billions of individuals possible, advances in data mining and Artificial Intelligence have enabled firms to learn more from data than was conceivable a few decades ago”2. This observation concerns big companies but small businesses are deprived of the capacity to collect and process these numerous forms of data. They need the assistance of large entities, like online platforms, which possess the technical and computing means to harvest the data and to exploit the information they convey. Hence, the specific service of Amazon, named “Amazon Custom”, puts its computing and Artificial Intelligence (AI) capacities at the disposal of its partners to help them to develop more personalised communication with their customers (“Prioritising personalisation of the user experience as a core growth objective”) and acquire a competitive edge over rivals.
3The use of Artificial Intelligence (AI) is becoming essential for mining and processing such an amount of data (Big Data), and for inferring an accurate profile of the customer from the numerous traces they leave on the Web. This paper explores some legal consequences of the customisation of the relation business-to-customer which is allowed or will soon be allowed with the help of AI. Up to now, AI has mainly been used for personalising the message (1) in order to better target the user and to catch her attention. It could be used in the future to personalise the product (2).
I. Personalisation of the communication
4Retaining the attention of persons in order to send them customised ads is the commercial game of the main actors of digital networks, mainly for online platforms that operate on “two-sided” markets, where the visitor is less regarded as a customer than a product3. Capturing the user’s attention and keeping her hooked maintains her within the same bubble and prevents her from visiting other sites. While efforts are made during the time of her visit to convert the visitor into a customer, specific ads are presented and lots of data are collected.
5Besides the targeting of advertising messages4, conducting machine learning on Big Data allows the technology firms and their partner retailers to better tailor their services to consumers5. The personalisation of the consumer experience, a marketing concept, results in specific recommendations and the customisation of the user interface, pretending to facilitate the customer experience. The pages that the user of a specific web site visualises differ from those that will be presented to another user.
6The interface design may be conceived to mislead the visitor and guide him towards choices that were not his. The customer could be the victim of a manipulative algorithmic system6. The law must put an end to the psychological manipulation and the misleading practices of “dark patterns”7. A new European regulation addressing that issue is being prepared, but already existing rules can also help solve the problem.
7The future Digital Service Act8 addresses some concerns related to recommender systems set up by big platforms : “Very large online platforms that use recommender systems shall set out in their terms and conditions, in a clear, accessible and easily comprehensible manner, the main parameters used in their recommender systems, as well as any options for the recipients of the service to modify or influence those main parameters that they may have made available, including at least one option which is not based on profiling…” (art. 29). Transparency and information will be required.
8Biased AI-assisted decisions should be prohibited according to the draft regulation on AI9 : “The prohibition covers practices that have a significant potential to manipulate persons through subliminal techniques beyond their consciousness or exploit vulnerabilities of specific vulnerable groups such as children or persons with disabilities in order to materially distort their behaviour in a manner that is likely to cause them or another person psychological or physical harm. Other manipulative or exploitative practices affecting adults that might be facilitated by AI systems could be covered by the existing data protection, consumer protection and digital service legislation that guarantee that natural persons are properly informed and have free choice not to be subject to profiling or other practices that might affect their behaviour.” Unfortunately the prohibition will cover only practices that cause psychological or physical harm; commercial or economic damage is not concerned. However, in its White Book on AI, the European Commission10, in line with its desire to create a “solid European regulatory framework for trustworthy AI”, stated that digital actors remain fully responsible for the compliance of AI with existing rules that protect consumers and govern the processing of personal data and that “any algorithmic exploitation of consumer behaviour in violation of existing rules shall be not permitted and violations shall be accordingly punished”.
9The European legislation, within the framework of the Artificial Intelligence Act, will soon submit “high-risk AI systems”, which could in particular generate an adverse impact on fundamental rights, including the protection of personal data and consumer protection11, to mandatory requirements of “transparency by design” and of information : “High-risk AI systems shall be designed and developed in such a way to ensure that their operation is sufficiently transparent to enable users to interpret the system’s output and use it appropriately” (art. 13.1). This is a perfect illustration of the compliance technique.
10Among the existing rules that could be used to fight these misleading practices, we should mention of course the protection of personal data12, rules concerning the liability for consumer privacy breaches, but also the prohibition of misleading commercial practices13 which cause or are likely to cause the average consumer to take a transactional decision that he would not have taken otherwise14. Article 7 of Directive 2005/29 states: “A commercial practice shall be regarded as misleading if, in its factual context, taking account of all its features and circumstances and the limitations of the communication medium, it omits material information that the average consumer needs, according to the context, to take an informed transactional decision and thereby causes or is likely to cause the average consumer to take a transactional decision that he would not have taken otherwise”.
II. Personalisation of the product
11Let us imagine that in the near future, the enhancement of the processing of high-dimensional datasets by a stronger AI will enable large firms not only to personalise the advertising message sent to the customer, but also to customise the product. Two levels of customisation could be reached: firstly the offer of merchandise that is specially manufactured according to the profile of the targeted customer; secondly, the offer of a product that could be personalised by the customer with the help of AI.
A. The “fast fashion” personalised
12The “fast fashion” system consists in the observation and the quick copy of the last trends of fashion and it is distinguished by the great reactivity of its production lines and distribution channels. The production of short runs, constantly renewed, allows industry (Zara, The Little Pretty Thing, …) to stick to the changes of the trends. So far, it has relied on information collected by its buyers and designers who attend fashion shows and trade fairs for clothing, and who read the fashion press, etc…
13Now, faster processing power of information will result from the exploitation of huge databases by machine learning. The changes in fashion trends will be detected much sooner by AI and even anticipated by machine learning. Moreover, the processing of data by AI will allow finer segmentation of the market and of potential buyers. It will allow better knowledge of the customer to be obtained and a better assessment of her needs and expectations. Finally, it will allow a more accurate profile of the targeted customer to be drawn.
14With that knowledge, the manufacturers of the clothing sector will be able to offer the customer a product corresponding to the latest in fashion and tailored to the personality of the potential buyer. “When they can identify individual tastes at fine levels and personalise their services to this taste, they often improve people’s lives” stressed the Stigler Committee15. Fashion victims will like that. A personalised “prêt-à-porter”16 of this type should be successful. It would give a strong competitive advantage over competitors, those who are deprived of access to the databases and AI resources.
15The manufacturer or the online platform which offers the services of its AI to a few clients could be accused of abusing a dominant position within the internal market by imposing unfair trading conditions, or by applying dissimilar conditions to equivalent transactions with other trading parties, thereby placing them at a competitive disadvantage17.
16Of course, the rules that have been mentioned earlier in this paper concerning the protection of personal data and the prohibition of misleading practices will apply to these new sales techniques.
B. Product as a Service (PaaS)
17The development of the use of AI in the distribution channels could generate a “fast fashion 2.0”, a combination of precise customer targeting and interactivity. To the quick capture of the fashion trends and the precise knowledge of the customer’s needs, tastes, expectations and financial means, could be added the possibility offered to the client to match her own colours, to arrange the components of the product in her own way, to choose the accessories, etc… within a predefined framework. Then the product is individualised. The result is a personalised product, tailored with the participation of the client and the help of AI. Once the client has made his choices, the product is immediately manufactured and soon delivered.
18The product is no longer a standardised commodity, produced and sold in multiple copies. It is unique, manufactured in accordance with the customer’s instructions. The product becomes a service (Product as a Service, PaaS). The legal consequences are important (nature of the agreement, consumer protection, consequences of defectiveness, etc…)
19Nature of the contract. Usually the contract between a retailer and a customer is deemed a sale. When the product is designed by the customer with the help of AI, and made by the manufacturer, is it a sale or a service contract? According to article 1787 of the French Civil Code, “when one gives a person a certain work to do, it may be agreed that he will furnish only his work or his industry, or that he will also furnish the materials”. In that case, the contract is still regarded as a service not a sale, provided that the product is not entirely designed in advance by the workman. If the characteristics of the product are not determined by the maker before the agreement, but are tailored with the purpose of matching the individual needs of the other party, it is not a sale. In the situation described in the previous section, the product is a service.
20Consumer protection. It is worth noticing that the consumer protection law does not make those subtle distinctions. It specifies that the contracts by which a professional offers the supply of goods to be manufactured or produced are deemed sales 18. Hence, the seller should check that the free choices left to the customer in the composition of the product do not jeopardise the fitness of the final product “for the purposes for which goods of the same type are normally used”19. Otherwise the seller will be liable to the consumer for any lack of conformity which exists at the time the goods were delivered.
21The protection of the consumer in her relation online with a trader is guaranteed by several provisions of directive 2011/83/EU and by the French Consumer Protection Code: provision of information before the consumer is bound by a contract, formal requirements for distance contracts, information on the price, right of withdrawal, etc… The withdrawal period expires 14 days from the day of conclusion of the contract, in the case of service contracts, and from the day the consumer acquires physical possession of the goods in the case of sales contracts.20 The nature of the contract, sale or service should not be neglected.
22On top of that, the PaaS could fall within the scope of an exception from the right of withdrawal. Article 16 (c) of the EU Directive states that the right of withdrawal should not be granted for the “supply of goods made to the consumer’s specifications or clearly personalised”. But courts have reduced the scope of this provision with the purpose of better protecting the consumer: the exception to the right of withdrawal only applies when the good is “clearly personalised” with regard to the specific needs of the customer21.
23In the case of a PaaS, even if the general sketch of the product is preset by the trader, the main features are chosen by the customer with the help of AI. The more freedom of choice the client has in designing the good, the more likely the exception to the right of withdrawal is to apply in the case of a PaaS.
24Liability for defective products. Should the liability of the producer for damage caused by the defectiveness of the PaaS be excluded? Pursuant to Directive 85/374/EEC22, liability without fault should apply to the producer, defined as “the manufacturer of a finished product” but also either as “the manufacturer of a component part” or as “any person who, by putting his name, trade mark or other distinguishing feature on the product presents himself as its producer”. The trader of a PaaS belongs to these categories. If the defectiveness of a product designed by collaboration before being manufactured by the trader causes damage, the latter will be held liable. But exonerating circumstances are mentioned by article 8.2 which offers the possibility to the producer to free himself from liability: “The liability of the producer may be reduced or disallowed when, having regard to all the circumstances, the damage is caused both by a defect in the product and by the fault of the injured person or any person for whom the injured person is responsible”. Is the participation of the customer in designing the product regarded as the equivalent to the “fault of the injured person”? Surely not, because the “fault of the victim” must be equal to a “force majeure”. The components and the overall structure of the good have been conceived by the trader with the help of AI. The same is true for the different combinations of elements left to the customer’s initiative. If the defectiveness of one of them causes damage, the liability of the trader should not be excluded.
25The developments described above may turn out to be purely imaginary or the economic model may turn out to be unprofitable and not very interesting to develop. But if these speculations come true, they will raise great concern. A core concern is the concentration of data, and economic and computational power in a few online platforms. The most worrisome topics are the skills they have acquired in developing interfaces that manipulate the minds of the consumers, and their outsized market power which enables them to act as private-rule makers23 and impose unfair conditions for smaller businesses eager to use their services. The solutions may lie in competition law.
Notes de bas de page
1 Proposal for a Regulation of the European Parliament and of the Council on a single market for digital services (Digital Services Act) and amending Directive 2000/31/EC, 15/12/2020, COM (2020) 825 final.
2 Stigler Committee on Digital Platforms, Final Report, Chicago, September 2019, Policy brief, p. 48.
3 CNIL, janv. 2019, Cahier Innovation & Prospective, n° 6, La forme des choix : Données personnelles, design et frictions désirables, p. 12 ; Ch. Dejean-Ozanne, Economie de la donnée et plateforme en ligne, PhD dissertation, Toulouse, 2019.
4 Digital Services Act, §52, 63.
5 Stigler Committee on Digital Platforms, p. 44.
6 The draft regulation Digital Services Act aims at ensuring the transparency of online advertising. Article 24 states that online platforms that display advertising on their interfaces must ensure that the recipients of the service can get “meaningful information about the main parameters used to determine the recipient to whom the advertisement is displayed”.
7 CNIL, janv. 2019, Cahier Innovation & Prospective, n° 6, glossaire; “user interfaces that make it difficult for users to express their actual preferences or that manipulate users into taking actions that do not comport with their preferences or expectations” (Stigler Committee, p. 8).
8 J. Larrieu, Ch. Le Stanc, P. Tréfigny, Panorama Droit du numérique, Rec. Dalloz 2021, p. 2158.
9 European Commission, Proposal for a Regulation of the European Parliament 2nd of the Council laying down harmonised rules on Artificial Intelligence (Artificial Intelligence Act, 21/04/2021, COM (2021) 206 final, §5.2.2 ; art. 5.1.a).
10 European Commission, White Paper on AI, 19 Feb. 2020, COM (2020) 65 final, 5-B.
11 Artificial Intelligence Act, §28.
12 Regulation (EU) 2016/679 of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, §24, 71, 72; art. 4. 4 ; Art. 82, Loi Informatique et Libertés n° 78-17du 6 janv. 1978.
13 Directive 2005/29 EC of 11 May 2005 concerning unfair business-to-consumer commercial practices in the internal market.
14 Dir., art. 6.1, 7 ; French Code de la consommation, art. L121-1, L121-3, al. 1 ; L121-4, 11°.
15 Stigler Committee report, p. 48.
16 Ready-to-wear.
17 Art. 102, TFEU.
18 Code de la consommation, art. 217-1 ; Dir. EU 1999/44, art. 1.4.
19 Dir. EU 19990/44, art. 2.2.c).
20 Dir. EU 19990/44, art. 9 ; Code de la consommation, art. 121-18, c.
21 Cass. Civ. 1, 17 janv. 2018, n° 17-10255.
22 Council Directive 85/374/EEC of 25 July 1985 on the approximation of the laws, regulations and administrative provisions of the Member States concerning liability for defective products.
23 Autorité de la concurrence, 12 July 2021, n° 21-D-17, Google : D. 2021, 1591, A. MENDOZA-CAMINADE ; 1624, J. LARRIEU.
Auteur
Professor emeritus
Université Toulouse Capitole-CDA
Le texte seul est utilisable sous licence Licence OpenEdition Books. Les autres éléments (illustrations, fichiers annexes importés) sont « Tous droits réservés », sauf mention contraire.
Qu’en est-il du code du commerce 200 ans après ?
États des lieux et projections
Corinne Saint-Alary-Houin (dir.)
2008
Qu'en est-il de la simplification du droit ?
Frédérique Rueda et Jacqueline Pousson-Petit (dir.)
2010
La réorientation européenne de la TVA à la suite du renoncement au régime définitif
Francis Querol (dir.)
2014
Regards critiques sur quelques (r)évolutions récentes du droit
Tome 1 : Bilans et Tome 2 : Réformes-Révolutions
Maryvonne Hecquard-Théron et Jacques Krynen (dir.)
2005