The Fallacy of Informed Consent: Linguistic Markers of Assent and Contractual Design in Some E-User Agreements

Orthodox contract law theory assumes that parties agree to the terms of a contract before entering into an agreement. However, recent factual evidence points towards the fact that consumers do not systematically read, and thus become informed of, the terms of a contract. Academics are asking for mandatory frameworks to ensure that informed consent is indeed sought and given by parties to a contract. The present study looks into the user agreements of four online companies that provide a marketplace for the sale of goods or free provision of services by other sellers and/or users (Ebay, Tripadvisor, YouTube and Amazon). The aim is firstly to identify the lexical/textual markers and peri-textual features of agreement in order to highlight the fallacy of informed consent. Secondly, the paper lists textual and peri-textual alternative contractual design (here called counter-design) in online user agreement. In so doing, contractual design features are distinguished from nudges. Suggested counter-design features help make informed consent effective.


Introduction
The founding principle of contract law theory is that parties give an informed consent when entering into a contractual agreement. Indeed, the idea of informed consent appears at several stages of contract formation. The essential elements of a contract are generally summarized as agreement, consideration and intention to create legal relations (Furmston, 1981: 24-105). Agreement is the requirement that the offeror makes an offer to the offeree about the provision of goods and/or services under certain conditions (price or consideration, quality, etc.), and that the offeree agrees unconditionally to the offer, before communicating this agreement to the offeror. In the very first steps of contract formation, it is assumed that both parties agree on essential terms for the provision of goods and/or services (Atiyah, 1995: 56-79). Informed consent is also required with regard to another condition for a valid contract, intention: parties to a contract need to intend to be legally bound by the agreement they enter into otherwise the contract is void ab initio. Consent -and the capacity to consent-to the terms of the agreement is given by both parties and is essential to the validity of a contract. If this element is missing, then not only will the contract not be binding (thus not enforceable in court), but will be considered as having never been entered into (James, 1989: 275-282).
However, in practice, it is a well-known fact that informed consent is absent from most agreements consumers enter into. Most consumer sales are carried out with standard form contracts (hereafter SFC). On the sellers' part, this means that the contract form is drafted by or with the help of legal counsel and according to standard practice in the industry. On the consumers' side, this implies that they cannot change the terms they are asked to agree to. Contrary to another contract law theory myth, most contracts (and SFCs in particular) are not the outcome of a negotiation between parties. Consumers are asked to agree to non-negotiable terms. This is the first flaw in orthodox contract law theory (Becher, 2007: 119).
Moreover, there is also a practical flaw as consumers do not read standard contracts (Bar-Gill, 2012: 3). There are three main psychological reasons for the no-read phenomenon. The first is that consumers are given so much information they are unable to process it in the time available. The second is the herding effect. Consumers will tend to trust a sale agreement they know others have agreed to. The third is consumer overconfidence. Consumers tend to think that small type or unfavourable provisions making amends for unexpected events are unlikely to happen to them (Avgoulea, 2009: 440f;Ayres and Schwartz, 2014: 545f). 2 The present paper focuses on the way in which informed consent -or what can be understood as the reading presumption-displays itself in a legal document. Our research is not so much interested in demonstrating the no-read phenomenon, which has been established for some time now, nor in mapping the fallacy of informed consent as in identifying how these phenomena are displayed at a linguistic level. Thus the linguistic markers of consent need to be identified (section 2). However, in order to carry out our study successfully, presentation of hypothesis, use of concepts and methodology first need to be outlined (section 1). Once the markers of this fallacy of informed consent are highlighted, the paper looks into the alternative markers of consent (here called counter-design) which would be effective in inducing consumers into informed consent, and thus overturning the practical shortcomings of contract law theory (section 3).
Working on a limited corpus of online user agreements, the paper seeks to identify the linguistic markers of the fallacy of informed consent by establishing the existence of a discrepancy between linguistic (lexical and textual) markers, which seem to help user assent, and peri-textual features, which make effective assent impracticable. By offering some clear linguistic tools (whose effectiveness is discussed in behavioural economics), the study contributes to rebuilding a consistent contract law theory.

Concepts
In the present study, markers are neither to be understood as discourse markers nor as pragmatic markers, i.e. sequentially dependent elements (Schiffrin, 1996: 31) which structure texts (Juncker and Ziv, 1998: 4), but considered as observable text features (Condamines and Péry-Woodley, 2007: 2) which impact on contract comprehension. These tools or handles point towards two functions in the user agreements studied here: first, linguistic (lexical) features of assent, and second, text construction (visual features of the document, such as typography and layout) related to agreement.
Lexical markers of assent are lexical tools indicating assent in a given text. They can take any grammatical form, ranging from verbal and noun forms, to phrases (including verbal forms with adverbs and tautologies). Textual markers are to be found in the layout of the documents, including such features as headings (Nunberg, 1990), contrasting disposition and typography (Virbel, 1989). Therefore the study of textual markers looks into document structure more generally (Power, Scot and Bouayad-Agha, 2003), otherwise called text architecture (Virbel, 1989) or layout structure (Delin, Bateman and Allen, 2002).
The study also focuses on peri-textual elements that determine socio-linguistic features of assent. Taking into account peri-textual elements of contractual agreements is consistent with the linguistic theory initiated under Todorov's analysis of Bakhtin's work. Indeed Bakhtin is quoted as writing: Let us agree to use the familiar word situation for the three implied aspects of the extraverbal part of the utterance: the space and time of the enunciation ('where' and 'when'), the object or theme of the utterance (that 'of which' it is spoken); and the relations of the interlocutors to what is happening ('evaluation') (Todorov, 1984: 42, Todorov's emphases). Time and place (i.e. layout of the text) are relevant peri-textual features in the present research.

Genre and Corpus
The genre of the documents is of utmost importance to the assumptions that are made at the outset of the study and for the conclusions drawn from the discrepancies between lexical/textual markers and peri-textual features in the selected corpus. Indeed, the issue of genre is not limited to literary and linguistic studies (Todorov, 1984: 80-85;Firth, 1969) but is essential to anchor the lexical and textual phenomena studied in a particular linguistic usage and with a specific strategy (Condamines and Péry-Woodley, 2007: 16).
The corpus selected for the present study is made up of contracts referred to interchangeably as 'Terms of Service', 'Conditions of Use', 'Terms, Conditions and Notices', or 'User Agreements'. The Terms of Service of four online companies were selected on account of their major position in the e-commerce industry. The prominence of these players in the online market is important as it creates both consumer optimism 3 and the herding effect, thus reinforcing the grounds for the no-read phenomena. E-bay is an online auction hosting website, 4 Tripadvisor offers online travel services 5 which are content-generated by users, Amazon is an online shopping and content distribution company, 6 and YouTube is a video-sharing website. 7 All the documents are therefore legal documents that define the rules governing the use of web-browsing services.
The companies were also selected because they offer provision of free service, coupled or not with purchasing services. E-bay, Tripadvisor, Amazon or YouTube users can browse through any online information on the website (on goods, services, holidays, restaurants, videos), and thus use a free service without having to register. Since to register means to be asked to effectively click on the Terms of Sale to notify agreement, a registered user cannot subsequently claim not to have been given the opportunity to be acquainted with those Terms. As the selected websites do not require registration for the use of the browsing services, they are relevant to study the present type of consent.
Our analysis is determined by the genre of the documents studied. Indeed, in a contract, a term is an utterance, which will become performative on agreement by both parties. That is, on agreement, both parties will be asked to perform particular acts. Thus the performative nature of the lexical terms and textual construction of assent is essential to understanding the function of linguistic markers and peri-textual features in online user agreements.

Working assumptions
Assumptions flow from the legal nature of the document. First, as in standard contracts, the paper assumes that Conditions of Use contracts mention the way in which the agreement is made. Signature of a legal document, whose purpose is clearly stated, usually indicates assent. The example below shows a standard form of assent in the real economy: […]  In certain cases, stress is also laid on the nature of the assent given by the parties to the agreement: Signing this agreement implies full understanding of the above conditions and the rental agreement. 9 Thus a contract can lay additional stress on the manner in which consent is given ('signing') and the legal consequence of giving one's assent to the terms of the document ('implies full understanding').
This example shows that markers of agreement need to be found in the choice of both lexical and textual markers. These markers are seen in phrases and terms such as 'entering into this contract', 'agreement', 'understanding', which point towards the performative nature of the lexical terms in the document. However -and this is one of the elements of contracts which sets them aside from any other legal documents-more is required from parties to express their assent than just the existence of words on a paper (which are presumed to have been read); a traditional paper contract will require parties to physically sign the document. The text itself includes a specially designated space provided for signatures. Textual markers are thus of importance to our study.
Though the corpus selected shares features with traditional contracts, it is part of a sub-genre of its own. Indeed all the companies chosen in the present study belong to the virtual economy. Documents in the corpus are e-contracts. The study of online agreements shows how different a standard contract in the real economy is compared with one in the virtual economy. As a consequence assent cannot be given with a physical signature. E-commerce users are familiar with such buttons as those below: Physical signatures are reborn as digital signatures in the e-commerce industry. However, this entails clear textual differences between contracts in the traditional retail and service industry and those in the e-commerce sector. A physical signature is unique and personal to an individual. Copying someone's signature with the intention of inducing someone else to consider it as genuine is an offence qualified as forgery. 13 A digital signature is an act (clicking on a web button), which is not unique (everyone clicks on the same button) but rests on the assumption that it is personal (it is I, and not anybody else who is clicking on the web button).
Notwithstanding these features of agreement, specific to the real economy on the one hand, and to the virtual industry on the other, the corpus selected here challenges methods of assent. Indeed, agreement is neither given through physical signature, nor by a click on a web button. As the present study focuses on the free provision of information (as opposed to purchasing services), web wrap (or browser wrap) 14 acceptance is the way in which users mark assent to Conditions of Use. In Terms of Service agreements, the user is held to the terms even if he has not read or clicked on an agreement button to indicate assent (Schneider, 2014: 325-8). This well-mapped procedure is nonetheless a major departure from traditional contract law theory, whereby the parties need to indicate agreement to the terms by which they will be bound. One of the assumptions is that these differences can be read into the linguistic features of the agreements themselves. Indeed, the corpus displays a wide range of lexical and textual markers, as well as peri-textual features, of assent.

Linguistic Markers of Informed Consent
On account of the virtual nature of the relationship between consumers and companies in the online industry, a certain number of features, such as signature, used to indicate assent in real contracts cannot be used. User Agreements also differ from e-contracts per se, in that assent is not given through clicking on 'I agree' web buttons. In Terms of Service, it is the use of the site that binds the visitor to the Conditions (Schneider, 2014: 328).
Moreover, it must be borne in mind that the e-commerce contracts are SFCs (standard form contracts). The terms of the contract can be neither negotiated nor changed to purpose, and they impose roughly the same conditions of use from one company to another. These are the reasons why the law regulating the agreement cannot be considered as a relevant element to the present study. 15 The browser wrap and standard form features of the agreements in the corpus make it impossible to apply orthodox contract theory that makes voluntary and informed assent to the terms of the contract mandatory to ensure the document is legally binding.

Lexical Markers of Assent
The legal documents in the corpus share common lexical features. There are three grammatical categories which are used in the four Terms of Service agreements to indicate assent. The most common are verbal forms: 'agree', 'understand', 'grant', 'acknowledge', 'declare', 'warrant', 'represent', 'are responsible for', 'consent to', 'signify your agreement to', 'read', 'consent'. Modals are also used: 'will/will not', 'may/may not'. Sometimes assent is reinforced through the use of adverbs ('hereby declare', 'are solely responsible', 'forever waive', 'forever release'), tautologies ('understand and agree', 'affirm, represent and warrant'), and binomial expressions ('represent and warrant'). Terms of Use agreements also use some isolated noun forms such as 'upon your acceptance'.
The following table summarises the frequency of occurrence of the lexical markers in the corpus. The following graph helps to see the distribution of occurrences of verbal forms signifying assent: The predominance of modals (may/may not) is consistent with the literature on the use of modals in legal writing (Richard, 2008), whereby 'may' is used to mark acts which are required or banned. Another marker is also used: will/will not. Grammatically, 'will' is both a modal and an auxiliary. Indeed it is primarily used as an auxiliary and only secondly as a modal, and as such is somewhat different from the usual modals. In a legal context, generally, 'will' is used in oaths or affirmations to mark the speaker's assent to the terms of the oath (Richard, 2008). This importance of 'will' where 'shall' should have been used (Richard, 2008) can only be explained by the need to make the meaning of the document accessible to lay users. Indeed, this is an SFC in the e-commerce industry. In accordance both with the contract law theory of informed consent, and with the plain English movement, users are presumed to have very little to no knowledge of the law. Indeed followers of the plain English movement believe that legal documents should be written in a language which is as simple and clear as possible (Mellinkoff, 1953;Wydick, 1994;Steinberg, 1991). Its legal use being excepted, 'shall' is rarely used as a modal in everyday language. It is thus consistent both with the plain English movement and with the need for informed consent, that an equivalent and more common modal be used in a consumer SFC.
If the use of modals ('will/will not', 'may/may not') is excluded, the most recurrent occurrence is the verb 'agree', which unambiguously indicates assent. The presence of this lexical marker was to be expected. However, the predominance of verbal forms of assent over noun forms needs to be accounted for.
In more than half of the occurrences (48 occurrences) the verb 'agree' appears in 'you agree' sentences. Assent can be signified in noun forms also: 'agreement', 'acceptance'. These noun forms can be found in noun forms in sentences such as: We may also ask you to acknowledge your acceptance of the User Agreement through an electronic click-through. 16 [Y]ou signify your agreement to be bound by these conditions. 17 This Website is offered to you conditioned upon your acceptance without modification of any/all the terms. 18 However, the occurrence of these noun forms is minor in comparison with verbal forms. The first reason given to explain the predominance of verbal forms is the performative nature of the agreement. Indeed, verbal forms in the document are in the present tense. They point towards the performance of the act upon agreement. The second reason given is the need to communicate in clear and unambiguous language in the wake of the plain English movement. Nouns point towards abstract mental phenomena, whereas verbs point towards actions.
Contrary to traditional contracts where parties are mentioned by their functions, in e-commerce standard terms of use, the pronoun 'you' is used to refer to the consumer entering into the agreement with the online company. The example below shows an instance of a standard US agreement in the real economy where the names of the parties are referred to by their function: This contract is an agreement between {Renter}, who will be renting a house from {Owner}, who owns the house being rented […]. The owner has a right to enter the house with an advanced notice of 24 hours for any reason.
[…] The renter will make his or her best effort to keep the house in good condition. 19 {Name}, henceforth known as "Builder," and {Name}, henceforth known as "Client," […] Client agrees to the estimate provided by the Builder on {date}, with the following changes (to which the Builder has agreed) […]. 20 In standard Terms and Conditions contracts in the virtual economy, the corpus shows the consistent use of 'you'. The consumer, who is presumed to be the man in the street with only a layman's knowledge of the law, is addressed to as 'you'. This complies both with the theory of informed consent, and with the plain English movement.
However, compliance with the plain English movement does not seem to rank high with drafters when they use some categories of lexical markers to reinforce agreement. The markers used in this case are adverbs, tautologies and binomial expressions. The use of these markers needs to be accounted for and in most cases is inconsistent with the overall aim to communicate in easily accessible legal English (Richard, 2014). The binomial expression 'understand and agree' is not used in all documents. It is used once in YouTube's Terms of Use and three times in Tripadvisor's Terms. Similarly, the use of 'hereby', another legal archaism, is to be found only in the Terms and Conditions of these two e-commerce companies. Even though these legal documents are SFCs which share common features with other contracts in the corpus (headings, mode of address, contents, etc.), they differ in that they were drafted by different teams of lawyers. This difference could account for the more traditional legal terminology to be found in YouTube's and in Tripadvisor's Terms of Service.
Notwithstanding the exception of some antiquated legal adverbs and binomial expressions, the study of the lexical markers of assent in the corpus of e-commerce Terms of Use reveals a clear aim to promote informed assent to the terms of the contract. Indeed, in most instances, the language used follows the requirements of the plain English movement, avoiding legalese (complex legal terminology, noun forms) to help understanding. The lexical markers show that informed consent is clearly sought by drafters of SFCs.
This strategy is reinforced by the use of textual markers. Indeed some of them are embedded in the documents themselves and highlight those terms which users need to be acquainted with.

Textual Markers of Assent
In all the documents in the corpus there are parts of the document in capitals and/or in bold type. The example below is an instance of this: The textual markers are designed to attract the reader's attention, and thus satisfy the requirement of informed consent. Their impact is reinforced by lexical markers in the heading, which restates the need to become acquainted with the content of the section:

PLEASE READ THIS SECTION CAREFULLY. IT AFFECTS YOUR RIGHTS AND WILL HAVE A SUBSTANTIAL IMPACT ON HOW CLAIMS YOU AND EBAY HAVE AGAINST EACH OTHER ARE RESOLVED. 22
Bold and/or upper case is used only for certain categories of information. Amazon does not use capitals in its conditions of use. It only uses bold to highlight the need to read the terms and the way in which users are bound by these terms (use of Amazon services). The other three companies consistently use upper case in one or two sections in their Terms of Use which relate to warranty and liability disclaimers and/or limitations. One exception is a long provision in the Ebay user agreement which forbids class actions. This provision is headed with a caption highlighting its importance and the agreement itself starts in its first section with a paragraph in bold indicating that the document contains information relative to legal disputes.
The legal document uses textual markers (bold and upper case), as well as headings and summaries, to identify important information a reader skimming through the document should not miss. These textual markers point towards the fact that effective informed consent on the part of the user is sought by the seller and/or service provider on a certain number of key legal issues.
Peri-textual markers, however, tell another story. As the present Conditions of Use are virtual verbal forms of communication (utterances), the time when the agreement is (or is not presented) to the consumer, as well as the position of the agreement on the webpage (space), are relevant to determine whether assent is effectively sought.

Peri-Textual Features and Assent
In some SFCs in the online retail industry, it is not unusual for the consumer to receive the Terms of Agreement after the purchase of the goods. This practice obviously challenges the presumption of informed consent (Hillman, 2002: 743-4). 23 In our corpus, the delay in informing users of the Terms is not as excessive, as Terms are accessible prior to purchase. However, peri-textual features of assent show problems of access to the Terms, in the context of web wrap (or browser wrap) acceptance. As commentators have shown, both elements stand in full contradiction with the orthodox contract law theory of informed consent.
The point in time when the user is asked to read the Terms of Use is of essence. For Conditions, it would be reasonable to expect that users be asked to agree to the Terms on first accessing the website. This logical requirement is nonetheless inconsistent with practice as in no case is the user actually asked to read the terms and acknowledge having done so on accessing a website. In all the online companies selected, the consumer is required to find the link to the Terms of Service on the page he/she is browsing on and to click voluntarily on the link in order to have access to the Terms.
In the YouTube website, the link to the Terms of Use is to be found at the very bottom of the home page, on the lower left hand side corner ('Terms'). This does not appear to be unusual as all the other sites in our corpus do the same. In the home webpage of the Amazon.co.uk website, the link to the Conditions of Use of the website is to be found at the bottom of ('Conditions of Use and Sale'). The location of the redirect link to the Terms of Use means that, in practice, the user will browse the site without any knowledge of the Conditions of Use.
Phrases as the following which are found nonetheless in the text of the terms and conditions are therefore of little use if the consumer does not first click on the link to the conditions of use and read the whole of the document: 'Please read the Agreement carefully'. 24 'Please read these conditions carefully before using Amazon Services'. 25 'If you do not agree to any of these terms, the Google Privacy Policy, or the Community Guidelines, please do not use the Service'. 26 'Please read this section carefully '. 27 Such phrases reinforce the fallacy of informed consent. Anybody reading the Conditions of Service out of their context of use will conclude that the utmost was done to ensure that customers have read them. However, the accessibility and timing of the terms seen above and conditions of access to them, as discussed below, contradict the effectiveness of the consent that is given to them.
Lexical markers of assent run through the documents and give the impression that consent is sought and given. However, a closer look at the Conditions of Use shows that consent is given through use, and not through reading the full terms and acknowledging them with a click-through. Sentences as the following indicate the web wrap (or browser wrap) nature of acceptance in these types of agreements: By using or visiting the YouTube website or any YouTube products, software, data feeds, and services provided to you on, from, or through the YouTube website (collectively the "Service") you signify your agreement to (1) these terms and conditions (the "Terms of Service"), (2) Google's Privacy Policy, found at http://www.youtube.com/t/privacy and incorporated herein by reference, and (3) YouTube's Community Guidelines, found at http://www.youtube.com/t/community_guidelines and also incorporated herein by reference. If you do not agree to any of these terms, the Google Privacy Policy, or the Community Guidelines, please do not use the Service. 28 By accessing or using this Website in any manner, you agree to be bound by the Agreement and represent that you have read and understood its terms. 29 Your continued access or use of our Services constitutes your acceptance of the amended terms. 30 By using Amazon Services, you signify your agreement to be bound by these conditions. 31 It becomes very clear on reading the above extracts that use of the services provided by the website implies consent. Assent is not given through reading the legal document but through use. As assent cannot be given after the user has perused the contents of the agreement, and as the user will nonetheless be held bound to the Conditions of Use, this mode of assent runs counter to orthodox contract theory.
On account of the principle of implied assent, the user is not given the opportunity to agree to the Terms of Service. Moreover, peri-textual features of assent show that s/he is not even given adequate notice of the implied assent s/he is giving to the Terms. The study of peri-textual features (timing and conditions of access) exemplifies the absence of notification of Conditions of Use. Notwithstanding the no-read phenomenon, the peri-textual structure of retrieval of the agreement makes effective access, and thus consent, unlikely. Insufficient notice of terms and the absence of opportunity to assent to them make the principles regulating the operation of browse-wrap agreements incompatible with contract theory of informed consent (Robertson, 2003).

From the Fallacy of Assent towards Effective Consent: Linguistic Alternative Design
The study of lexical and textual markers of assent is evidence that legal texts, such as the standard User Agreement studied in the corpus, cannot be taken out of the context in which they are presented to the consumer for assent (peri-textual features). The lexical and textual phenomena are anchored in the features of certain contracts (SFC), in web usage (accessibility of terms on the website) and in the specific strategies that websites develop to give the linguistic impression that consent is sought in earnest, even if it cannot be effectively given in practice.
This is a double fallacy. The first level of deception is to be found in the legal document itself, which displays all the lexical and textual markers of assent as demonstrated above. These markers create a presumption of informed consent. This presumption weakens the judicial protection that could have been afforded to parties, who cannot then claim that they did not agree to the terms ( Bar-Gill and Ben-Sahar, 2012). In court, judges interpreting these Terms of Use would be bound to agree that consent was honestly sought, and that, if the user neither read nor understood the Conditions of Service, the online company cannot be held liable for such failure. The lexical and textual markers of assent clearly contribute to the misbelief that consent was lacking in the agreement the consumer involuntarily entered into, and thus the agreement should be null and void. This is indeed prejudicial to any claim on his/her part and may, as such, explain the insertion of these linguistic features in the Terms of Use, i.e. protecting companies from any claims based on the absence of the user's assent. Another reason may be the standard nature of these documents, since they are all SFCs. The first level of deception cannot be identified if the lexical and textual markers of assent are not put into the peri-textual context in which the Terms are presented to the user for agreement.
The second level of deception is embedded in the first. The textual markers of assent overshadow points the user did not expect to be bound by. Banking on the herding phenomenon and on consumer overconfidence, unexpected clauses are hidden in User Agreements. By highlighting certain clauses, others -which might have been of utmost importance to the consumer-are less likely to be read or identified by the e-consumer (Avgoulea, 2009: 12;Ayres and Schwartz, 2014: 551). Thus, by highlighting certain provisions, others are masked. The present section looks into the peri-textual and textual mandatory disclosure which could be introduced to make sure that consent is truly sought, and that the user has effectively read and understood the Conditions of Use.

Designing Effective Consent
The peri-textual and textual amendments suggested to improve the no-read phenomenon are part of contract distortions implemented with the aim of carefully designing environments which will make consumer choice of certain options more likely. This environment design is also known as choice architecture (Sunstein, 2012: 6;Sunstein, 2014: 14) and in this respect, any individual who seeks to modify an environment to make certain types of choices more predictable is known as a choice architect.
Nudgers are one type of choice architects. They voluntarily seek to organise an environment to maximise choice according to certain goals (political, health-related, economic, etc.), which are benevolent and, at the same time, cost minimizing. Sunstein (2014: 17) defines nudges as "initiatives that maintain freedom of choice while also steering people's decisions in the right direction". Nudges are thus initiatives operating in the interest of the nudgee. Sunstein uses the word 'right' in his definition. It is the nudger who determines the right goals for nudgees. The goals are therefore right only in the eye of the nudger, who may have a very different agenda from the nudgee. In this respect, a nudger is a paternalist. 32 Nudges are paternalists as they are initiatives devised with a benevolent aim and with the welfare of the nudgee at heart. However, not all architectural design initiatives aim at benefitting the individual. Some aim at benefitting the financial interests of a company. In a business-related environment, as the one selected for our corpus, the presumption is that the choice-environment will not be designed to enhance consumer well-being, but rather business interests. Nudges are thus less likely to exist in this type of environment. Purists, consequently, would tend to refer to the designs found in our corpus as choice architecture initiatives.
Cass Sunstein explains that a "choice architect has the responsibility for organizing the context in which people make decisions" (Sunstein, 2012: 6;Sunstein, 2014: 14). Some disclosure mandates aim to alter people's behaviour to induce them to read parts, if not all, of the agreement they enter into. Disclosure mandates have the advantage of being based on a psychology-related scientific method which looks into the reasons why individuals do not read contracts (quantity of information, herding effect, consumer overconfidence, etc.). They devise initiatives (click-wrap agreements, ordering information, and warning boxes) that seek to make the response of individuals predictable.
It is nonetheless important to note that users will not be forced to modify their behaviour and read the document. Even if it may be easier to become aware of a certain number of provisions they are bound to, consumers will still have the choice to select another option, that is, to refuse to read the information provided. Thus, a characteristic feature of the present initiative is that its reversal is possible, and thus the user may override the choice architect's preference (Sunstein, 2014: 17). This feature may, at first sight, be seen as sharing some common grounds with nudges. As A-L Sibony and F. Alemanno note, "[n]udge is […] presented as a distinctive, alternative way, characterized as being minimally burdensome, low-cost and choice-preserving, to help promote regulatory goals" (2015: 7). However, as there is no alternative available for any user turning down the Terms, the initiatives studied fail to be nudges on account of the absence of effective reversal of the initiative. If consumers want to make secondhand purchases on the web or if they want to view videos, E-bay and YouTube have a quasi-monopoly on the market. Refusing to comply with the Conditions implies that consumers are barred from access to a service of the same quality on the web. Moreover it is likely that other competitors impose the same conditions. The disclosure mandates will not guarantee that Terms are indeed read, but they will give consumers a fair and time-efficient opportunity to become aware of key (or unexpected) features of the agreement, even if, once aware of the terms to which they are bound, they can neither amend them, nor turn them down, as there is no real alternative on the market.
When it comes to initiatives operating in a business-related environment, the presumption of benevolence is rebutted, as the initiative is likely to enhance business profits at the expense of effective consumer choice. The textual and peri-textual features of the legal documents in the corpus can now be re-read as architectural designs aimed at avoiding actual consumer consent on purpose. The only way to counter consumer misconceptions, which are exploited by sellers and service providers to design their User Agreement structures, is to construct alternative contractual designs (counterdesign) to counter the effect of the existing contract designs. Click-wrap agreements and disclosure mandates are construed as initiatives countering design features (Bar-Gill, 2012 Deconstructing contract law principles as they operate in practice is a sterile academic endeavour if it is not coupled with the aim of offering ways in which everyday online agreements can be made compatible with contract law theory. Some legal scholars have argued that browse-wrap agreements should be replaced by click-wrap agreements, whereby the user needs to click that s/he has read and agreed to the terms of use (Robertson, 2003). Web users are familiar with click-wrap agreements such as follows: In theory, the suggestion would indeed help notify users of the existence of certain terms regulating the website and of the Conditions of Use themselves through change of the peri-textual features of assent (timing and access). In practice, there is evidence that Courts (in US jurisdictions) are more likely to uphold click-wrap agreements than browse-wrap agreements. Indeed assent to the former is construed as effective, since the opportunity was given to review the agreement and there is the obligation to click-agree to the conditions. 34 If the click-wrap agreement suggestion made it possible to improve the opportunity to consent, it would not guarantee effective consent in practice. One of the reasons for this failure is that it does not take into account the no-read phenomenon. Even if consumers were given the opportunity to click-wrap Service Conditions, it would not entail the actual obligation of reading through the Terms themselves. In theory, the principle of contractual assent would be enforced, but in practice, informed consent would almost never be given as users do not read and rarely scroll down click-wrap pop-up windows.

Overturning Shortcomings of Contract Law Theory: Textual Organisational Counter-Design
To improve effective assent and uphold contract principles in practice, scholars need to tackle the no-read issue. For almost half a century, the no-read issue has created problems for academics (Lewellyn, 1960;Kessler, 1943;Rakiff, 1983). Lewellyn (1960: 370) famously established the blanket assent theory to justify the enforcement of unread terms. According to that theory, a consumer may give his assent to a reasonable term in an SFC, but will not be held bound by terms deemed unfair.
This approach is of interest for our present study as it introduces a hierarchy of implied consent according to the reasonableness of terms. This reasonableness can be read at different levels. Reasonableness can be a prescriptive standard. A term would be considered as reasonable on account of its being fair in the contractual bargain between parties. Reasonableness can also be a descriptive standard. A term would be considered as reasonable on account of its being generally found in SFCs. An unreasonable term would be an unexpected term.
A certain number of academics have tried to find solutions to the no-read problem by investigating the issue of unexpected terms in contracts (Avgoulea, 2009;Ayres and Schwartz, 2014). This has allowed a more subtle approach to the disclosure strategy, as a means to fight the no-read phenomenon. The focus has shifted from mandatory disclosure regulation, which is costly and prices out poorer consumers (Bar-Gill and Ben-Shahar, 2012), to more optimal disclosure mandates (Bar-Gill, 2012). According to the latter approach, to determine which terms would be unexpected in a contract, legal scholars look into whether consumers held accurate beliefs as to the agreements they were entering into. This term substantiation would be necessary to determine which terms need to be highlighted for the consumer (Ayres and Schwartz, 2014). Only information statistically important for consumers would be disclosed (Bar-Gill, 2012).
In the present corpus, all Service Agreements contain textual markers to highlight certain provisions that are important for the seller or the service provider. The issue is to determine whether the same terms are of relevance to consumers. The SFCs studied here highlight sections in their Terms of Use which relate to warranty and liability disclaimers and/or limitations. Ebay has a long provision in its User Agreement forbidding class actions. Consumers are familiar with User Terms and would expect there to be disclaimers and provisions for preference forums in case of litigation. However, consumers might not be aware of the fact that the rights to any review posted on forums are transferred to the website: When providing us with content or causing content to be posted using our Services, you grant us a non-exclusive, worldwide, perpetual, irrevocable, royalty-free, sublicensable (through multiple tiers) right to exercise any and all copyright, publicity, trademark, and database rights and other intellectual property rights you have in the content, in any media known now or developed in the future. 35 This is an unexpected feature common to all four Conditions of Use in the corpus, a feature which consumers should be made aware of. One way of making sure that consumers read this provision is either to have the information inserted at the beginning of the agreement, or to have warning boxes (Ayres and Schwartz, 2014). The idea underlying the approach of some academics is that if you diminish the amount of irrelevant information, you will increase the likelihood of consumers reading the documents (Avgoulea, 2009). Once the amount of information is diminished, the textual order in which it is presented is of essence (Ayres and Schwartz, 2014). It is indeed a fact established by psychological evidence that assent can be manipulated by the framing of information (Avgoulea, 2009). Information thus needs to be organised in descending order according to consumer importance. If one considers the picture of the click-wrap agreement given above, the important or unexpected information should be visible in the box and thus could easily be read or skimmed through by the consumer, without having to scroll down.
The ordering of information and the textual features of warning boxes should, in principle, be efficient counter-designs features to offer consumers an effective opportunity to agree to the Terms.

Conclusion
Working on a corpus of online User Agreements, the present study has offered a twostep critical approach to the mandatory condition of assent in orthodox contract theory. Firstly, the study of the lexical and textual markers of assent in the corpus has highlighted that assent is indeed sought. However, on account of the no-read phenomenon, consent is not effectively given. This absence of effective assent is reinforced by the study of peri-textual features which show that assent is not effectively sought. The first part of the study uncovered the fallacy of consent in online User Agreements. The second part of the study has sought to reconcile real online user agreements by suggesting amendments to the textual and peri-textual features to make assent real, through the use of counter-design.
All around the world, academics, but also politicians and journalists join in to try to work out an assessment of choice architecture. 36 The present paper contributes to the debate in presenting a linguistic approach on these new regulatory instruments, which stand at the cross-roads of law, psychology and behavioural economics. Counterdesigning is one of the effective (although not compulsory) ways in which the no-read phenomenon can be tackled, thus ensuring true consent for agreements entered into by consumers. In the corpus studied, design will not be found in redrafting the Terms of the Service but in reorganising peri-textual and textual features of the agreements. The linguistic forms of design have thus been highlighted in standard forms of user-wrap agreements of online companies.
The linguistic approach to contract law theory has helped to identify non-legal features (click-wrap agreements, ordering information, and warning boxes) that need careful counter-designing to guarantee theoretical validity of contractual agreements. Notes 1. A draft of this paper was presented at the École de Droit (Sciences Po, Paris) Nudge Seminar on Friday 27th November 2015, at Assas University. I would like express my gratitude here to the anonymous reviewers of this article and to the French Nudge Project team (http://frenchnudgeproject.fr) for their helpful remarks and suggestions.
2. Behavioural economics studies consumer systematic misperceptions to identify behavioural market failures (Bar-Gill, 2012;Becher, 2007;Avgoulea, 2009). These misconceptions can affect perception of product features and use patterns of products (Bar-Gill, 2012). At the time of the assent, psychological studies highlight flaws in the perceptions consumers have of their commitment to the terms. Among these flaws will be sunk-cost effects, cognitive dissonance, confirmation bias, low-ball technique, low-probability risk, availability cascade and self-serving biases (Becher, 2007). Behavioural economic studies make it possible to think outside the standard rational choice theory, to take into account predictable consumer bias which affects choice (Bar-Gill, 2012).
3. This concept from the field of behavioural economics is synonymous with consumer overconfidence.
14. Browse-wrap agreements are usually accessible by clicking on a hyperlink, thus users are not required to affirmatively agree to the terms. By contrast, click-wrap agreements ask