# Propositions from Contracts in the Age of Smart Readers

**Citation:** Yonathan A. Arbel & Shmuel I. Becher, Contracts in the Age of Smart Readers, 90 Geo. Wash. L. Rev. 83 (2022)

**Source:** [final published PDF](https://www.gwlr.org/wp-content/uploads/2022/02/90-Geo.-Wash.-L.-Rev.-83.pdf)

**Review status:** 18 model-drafted, source-checked; 0 human-reviewed. Page references use the printed pagination and, separately, the 1-based PDF page number.

## 1. Language-model smart readers can change consumer contracting by simplifying, personalizing, constructing, and benchmarking boilerplate

**Location:** Introduction and Part I, Smart Readers: Technology and Capabilities, printed pp. 83-94 (PDF pp. 1-12)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 83–94, that language models can become inexpensive, widely accessible smart readers of contracts, disclosures, and privacy policies. They organize the technology around four capabilities: translating difficult text into usable explanations, personalizing presentation to the reader, constructing legal consequences from terms, and benchmarking one contract against market alternatives. This is significant because contract law has long treated unread boilerplate as a stubborn human and institutional problem, while smart readers recast at least part of it as a technological problem whose solution can reshape assent, competition, and regulation. It connects to informed consent, consumer protection, legal automation, natural-language processing, boilerplate design, access to justice, and the allocation of responsibility when an automated explanation is wrong.

**Evidence anchor:** The abstract and introduction define smart readers, preview four capabilities and their benefits and risks, demonstrate early outputs for difficult clauses, and frame the article as a forward-looking inquiry into adoption, market consequences, and legal response.

**Boundary:** The article uses early GPT-3 outputs selected to illustrate future capabilities, expressly acknowledges errors and weak reliability, and evaluates smart readers against realistic alternatives such as nonreading rather than against perfect legal advice.

**Connections:** consumer boilerplate; informed assent; natural-language processing; legal automation; consumer protection; access to justice

**Record:** `ssrn-3740356-p01` · `machine-drafted-source-checked`

## 2. Smart readers can make dense contracts accessible through more than mere shortening, but simplification necessarily risks losing meaning

**Location:** Part I.A, Simplification, printed pp. 95-99 (PDF pp. 13-17)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 95–99, that smart readers can reduce contractual complexity by summarizing, lowering the language register, shortening and restructuring sentences, changing formatting, removing nonessential material, and adding examples or a more engaging voice. Their lease and at-will-employment examples show that an explanation can be easier to process even when it is not shorter than the original. This is significant because the proposal addresses several sources of unreadability at once instead of assuming that a shorter disclosure is automatically a comprehensible one. It connects to plain-language drafting, disclosure design, cognitive load, legalese, contractual bloat, and lossy compression, while the model’s serious legal mistake about discrimination illustrates why accessibility and accuracy must be evaluated separately.

**Evidence anchor:** The section identifies semantic difficulty, length, formatting, and legalese as barriers, compares original lease and employment clauses with model explanations, and characterizes smart-reader simplification as liberal, accessible, and necessarily lossy.

**Boundary:** Summarization inevitably omits information; some contractual length serves precision; and the demonstrated model produced a major legal error, so readable output is not necessarily reliable output.

**Connections:** plain language; contract design; cognitive load; legalese; lossy compression; AI hallucination

**Record:** `ssrn-3740356-p02` · `machine-drafted-source-checked`

## 3. Consumer-side personalization can adapt a uniform contract to a reader’s language, cognition, culture, and intersecting characteristics without requiring the firm to know each consumer

**Location:** Part I.B, Personalization, printed pp. 99-104 (PDF pp. 17-22)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 99–104, that the one-size-fits-all disclosure model fails because readers differ in language, culture, idiom, cognitive ability, socioeconomic circumstances, and preference for abstract or concrete explanation. A smart reader can personalize the presentation on the consumer’s device—translating and simplifying for a recent immigrant, explaining through examples for a teenager, adapting regional language, or combining several characteristics—without requiring the seller to collect the same information. This is significant because it moves personalization from the drafter’s side, where it is expensive and potentially exploitative, to the reader’s side, where it can serve comprehension. It connects to linguistic access, disability and cognitive accommodation, intersectionality, private dictionaries, the reasonable-consumer standard, and the difference between consumer-serving and seller-serving personalization.

**Evidence anchor:** The authors use Spanish-language, youth-oriented, concrete-example, regional-dialect, and intersectional examples to show how smart readers can tailor output and move parties toward shared dictionaries.

**Boundary:** Personalization does not reveal the substantive content of governing law, remains lossy, cannot completely eliminate divergent private meanings, and later sections show that personalization can also facilitate discrimination.

**Connections:** personalized disclosure; linguistic access; intersectionality; private dictionaries; reasonable consumer; consumer-side technology

**Record:** `ssrn-3740356-p03` · `machine-drafted-source-checked`

## 4. Smart readers can sometimes explain the legal consequences of simple terms, although their construction cannot be authoritative and may implicate unauthorized-practice rules

**Location:** Part I.C, Construction, printed pp. 104-106 (PDF pp. 22-24)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 104–106, that understanding a contract requires more than parsing its words: readers often need construction of a term’s legal effect. They show a smart reader explaining the consequences of buying a car “as is” and answering whether brief foreign travel triggers a permanent-residence default clause. This is significant because it positions the technology not merely as a readability aid but as a low-cost source of preliminary legal orientation for ordinary questions. It connects to the interpretation-construction distinction, consumer legal education, follow-up questioning, unauthorized practice of law, and the comparative baseline of what a reasonable lawyer—or an unaided consumer—would provide.

**Evidence anchor:** The section distinguishes construction from linguistic interpretation, tests the model on an as-is sale and residence clause, and cautions that legal disagreement and unauthorized-practice concerns constrain this capability.

**Boundary:** Construction is contested even among lawyers and judges; outputs are nonauthoritative, can omit defenses and state-law variation, and should be judged differently for mundane and penumbral questions.

**Connections:** contract interpretation; legal construction; consumer legal education; unauthorized practice of law; interactive legal tools; legal consequences

**Record:** `ssrn-3740356-p04` · `machine-drafted-source-checked`

## 5. Benchmarking can reduce comparison costs by scoring contract terms against the market and directing consumers to better alternatives

**Location:** Part I.D, Benchmarking, printed pp. 106-109 (PDF pp. 24-27)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 106–109, that benchmarking may be the most powerful smart-reader capability because it can score a contract relative to market practice, explain the score, and point the consumer toward sellers with better terms. Privacy-policy classifiers and PrivacyCheck illustrate how a model can identify clauses, compare an agreement to its sector, and compress a difficult choice into a familiar ranking. This is significant because consumers need not master every clause for contract quality to become a salient product attribute. It connects to comparison shopping, search costs, choice overload, reputation systems, privacy nutrition labels, term competition, and the accumulation of contract corpora that can improve market-specific comparisons.

**Evidence anchor:** The authors discuss machine classification of privacy provisions, a deployed browser tool that scores policies against market averages, explanatory rankings, competitor links, and the practical value of imperfect scores.

**Boundary:** Contract scoring is nascent and contestable, depends on normative judgments and relevant comparators, and cannot yet rival a seasoned lawyer; its value rests on improvement over intuitive nonreading, not perfect accuracy.

**Connections:** contract benchmarking; comparison shopping; search costs; choice overload; privacy policies; reputation scores

**Record:** `ssrn-3740356-p05` · `machine-drafted-source-checked`

## 6. Observed adoption of smart readers can discriminate among competing explanations for why consumers do not read contracts

**Location:** Part II, Smart Reader Uptake and (No) Reading Theories, printed pp. 109-114 (PDF pp. 27-32)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 109–114, that smart-reader uptake will depend not only on price and interface but on why consumers currently avoid contracts. Readability theory predicts strong adoption; transactional-expectations theory predicts use mainly in unfamiliar or high-stakes domains; rational-apathy theory predicts uptake when automated review materially lowers cost; cognitive-bias theories predict selective use by consumers aware of their limitations; and social-norm theories predict more private than face-to-face use. This is significant because adoption becomes a Wittgenstein’s-ruler test: a tool that removes the barrier posited by a theory also supplies evidence about whether that theory was sound. It connects to empirical consumer behavior, the privacy paradox, rational ignorance, transactional norms, behavioral bias, technology adoption, and the design of subsidies or interfaces.

**Evidence anchor:** Part II maps five theories of nonreading to distinct uptake predictions, discusses technical and economic feasibility, and explains how actual adoption can test the causal accounts behind consumer-contract policy.

**Boundary:** The authors do not predict a single adoption rate; effectiveness, cost, business model, conflicts of interest, UI/UX, domain, social setting, and consumer heterogeneity remain unresolved empirical variables.

**Connections:** no-reading theories; technology adoption; rational apathy; transactional expectations; cognitive bias; social norms; Wittgenstein's ruler

**Record:** `ssrn-3740356-p06` · `machine-drafted-source-checked`

## 7. Modest use of imperfect smart readers can improve individual matching and generate market-wide pressure for better contract terms

**Location:** Part III.A, Matching, Search Costs, and Market Competition, printed pp. 114-118 (PDF pp. 32-36)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 114–118, that greater term transparency has both micro and macro effects. Individually, consumers can find contracts better matched to their preferences and reduce search costs; collectively, even a minority that reads simplified terms or checks scores may exert enough demand pressure to improve standardized terms for everyone. This is significant because smart readers can revive a modest informed-minority theory without requiring most consumers to read contracts in full. It connects to product-attribute competition, shrouded terms, privacy markets, market entry, consumer activism, watchdog journalism, regulatory supervision, reputational pressure, and the spotlight effect on contract drafters.

**Evidence anchor:** The section connects transparency to matching and welfare, explains the informed-minority mechanism, works through privacy-policy and search-engine examples, and adds watchdog, agency, reputational, and moral channels of pressure.

**Boundary:** The dynamics depend on consumers caring about terms and on some mechanism for demand, entry, advocacy, or regulation; unequal access and firms’ ability to identify informed users can undermine the spillover.

**Connections:** informed minority; term competition; search costs; consumer matching; market entry; consumer activism; positive spillovers

**Record:** `ssrn-3740356-p07` · `machine-drafted-source-checked`

## 8. The most serious smart-reader risks arise from correlated error and deliberate adversarial manipulation, not simply from isolated mistakes

**Location:** Part III.B, Errors and Adversarial Attacks, printed pp. 118-124 (PDF pp. 36-42)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 118–124, that error must be evaluated comparatively and by type. Random errors may cancel in large markets, and an inexpensive, consistent tool can help even when it underperforms a lawyer because the realistic alternative is often no reading; correlated errors can systematically distort particular clauses; and adversarial examples let firms subtly alter text or formatting so a machine produces a seller-favorable interpretation invisible to a human reader. This is significant because the black-box reader creates a new strategic drafting surface on which apparent contract language and machine-perceived language can diverge. It connects to machine-learning security, correlated bias, platform contract scores, adversarial examples, conspicuousness doctrine, ALL-CAPS disclosures, strategic boilerplate, and the difference between replacing nonreading and replacing legal counsel.

**Evidence anchor:** The section compares human and machine performance, distinguishes isolated, correlated, and adversarial errors, demonstrates hidden-text and visual attacks, and analogizes machine manipulation to courts’ slow response to ineffective ALL-CAPS drafting.

**Boundary:** The authors expect substantial near-term error and acknowledge that their illustrations are cherry-picked; gradual adoption may bound harm, but sophisticated attacks can transfer across models and be extremely difficult to distinguish from innocent design choices.

**Connections:** adversarial machine learning; correlated error; black-box systems; conspicuous disclosure; strategic drafting; comparative accuracy

**Record:** `ssrn-3740356-p08` · `machine-drafted-source-checked`

## 9. Low-cost smart readers can scale basic know-your-rights assistance where subsidized human legal services cannot

**Location:** Part III.C, Access to Justice, printed pp. 124-126 (PDF pp. 42-44)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 124–126, that smart readers can relieve part of the access-to-justice deficit by providing on-demand explanations of contractual rights. A consumer like Ms. Williams, confronted with an obscure cross-collateral clause, could receive a warning, an explanation, and perhaps a referral to better alternatives on a device she already owns. This is significant because human legal subsidies face severe scaling limits, whereas software can distribute preliminary legal information at low marginal cost to people excluded by price, geography, and repeat-player advantage. It connects to legal deserts, legal aid, know-your-rights tools, smartphone access, repeat-player theory, consumer education, and the use of automation to complement rather than simply replace lawyers.

**Evidence anchor:** The authors describe cost, repeat-player, rural, and social barriers; explain why lawyer subsidies do not scale; and revisit the Williams cross-collateral example to illustrate low-cost, on-demand assistance.

**Boundary:** Smart readers are unlikely to match lawyers in the short or medium term; smartphone ownership does not eliminate the digital divide; and automated advice can itself contain bias or error.

**Connections:** access to justice; legal deserts; legal aid; know-your-rights services; repeat players; digital inclusion

**Record:** `ssrn-3740356-p09` · `machine-drafted-source-checked`

## 10. Better contractual awareness can reduce accidental breach but can also induce harmful compliance with illegal or unenforceable terms

**Location:** Part III.D, Compliance and Overcompliance, printed pp. 126-127 (PDF pp. 44-45)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 126–127, that clearer awareness, comprehension, and recall can improve compliance by both consumers and sellers, reduce accidental breach, and help consumers invoke promises such as repair or cancellation rights. The same clarity can be harmful when readers assume every written term is valid and morally binding, because form contracts often contain illegal or unenforceable provisions. This is significant because it rejects the simple premise that more reading is always pro-consumer. It connects to unenforceable noncompetes, exculpatory terms, chilling effects, the moral psychology of promise, legal literacy, disclosure policy, and the difference between identifying contractual text and assessing its enforceability.

**Evidence anchor:** The section identifies compliance benefits, then draws on studies of noncompetes, exculpatory clauses, and lay beliefs to argue that readable terms can acquire excessive legitimacy and deter valid claims.

**Boundary:** The direction of the effect depends on whether the tool can contextualize enforceability and remedies; improved recall is beneficial for valid obligations but potentially harmful for vulnerable clauses.

**Connections:** contract compliance; overcompliance; unenforceable terms; moral obligation; legal literacy; disclosure effects

**Record:** `ssrn-3740356-p10` · `machine-drafted-source-checked`

## 11. Smart readers can expose discriminatory contract personalization while also enabling firms to discriminate between users and nonusers

**Location:** Part III.E, Discrimination and Personalization, printed pp. 127-131 (PDF pp. 45-49)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 127–131, that personalization has opposite distributive possibilities depending on who controls it. A consumer-side reader can flag unusually harsh interest rates, tailor explanations to people excluded by the imagined white, educated, male reasonable consumer, and make intersectional accommodation feasible. But firms may identify likely smart-reader users, give them favorable terms, and finance those benefits through worse terms for less informed or digitally excluded consumers. This is significant because a technology capable of resisting individualized exploitation can itself become a basis for market segmentation and regressive cross-subsidy. It connects to algorithmic redlining, proxy discrimination, reasonable-consumer doctrine, big-data scoring, digital inclusion, disparate treatment, intersectionality, and the informed-minority assumption that firms cannot identify who is informed.

**Evidence anchor:** The section contrasts seller-side targeting with consumer-side accommodation, explains benchmarking’s protective potential, and develops a separating-equilibrium risk based on firms’ growing ability to score and identify consumers.

**Boundary:** Benchmarking needs an appropriate comparison group, personalized markets make comparators harder to identify, individual alerts cannot solve systemic discrimination, and mimicry is unavailable to consumers facing deeper digital barriers.

**Connections:** algorithmic discrimination; contract personalization; digital divide; reasonable consumer; market segmentation; regressive cross-subsidy; intersectionality

**Record:** `ssrn-3740356-p11` · `machine-drafted-source-checked`

## 12. Smart readers create a new channel for countering cognitive overload, risk myopia, and price manipulation at the moment of contracting

**Location:** Part III.F, Nudging with Smart Readers, printed pp. 131-133 (PDF pp. 49-51)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 131–133, that smart readers can function as consumer-side nudges against several recurring decision failures. They can reduce cognitive overload through summaries, scores, and accessible formatting; counter optimism and myopia by making warranties, return policies, and future risks salient; and defeat partitioned or psychologically manipulative pricing by calculating and presenting a rounded total transaction price. This is significant because the technology supplies a personalized intervention channel at the point when boilerplate and pricing architecture shape choice. It connects to behavioral law and economics, salience, cognitive overload, smart disclosure, choice architecture, drip pricing, left-digit effects, and field experimentation on consumer debiasing.

**Evidence anchor:** The section identifies three target problems, explains how summaries and formatting reduce overload, how salience can counter optimism and myopia, and how whole-transaction calculations can expose price partitioning.

**Boundary:** The tool cannot address every bias, and whether these interventions actually improve decisions is an empirical question that requires testing by researchers and consumer organizations.

**Connections:** behavioral nudges; cognitive overload; risk salience; partitioned pricing; choice architecture; smart disclosure

**Record:** `ssrn-3740356-p12` · `machine-drafted-source-checked`

## 13. If smart readers materially solve nonreading, consumer-contract interventions cannot continue to rely on information failure without reexamining their justification

**Location:** Part IV.A, The Challenge to Consumer Protection, printed pp. 133-136 (PDF pp. 51-54)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 133–136, that lack of reading may increasingly become a technological challenge rather than an immutable ethical premise for legal intervention. Because the no-reading problem supports positions across the debate over the Restatement of Consumer Contracts—including relaxed formation, unconscionability review, and skepticism toward merger clauses—effective smart readers could make some rationales dated and redirect policy toward adoption. This is significant because it asks whether consumer law is future-proof when a foundational account of informational asymmetry changes. It connects to meaningful assent, merger clauses, unconscionability, mandatory disclosure, the Restatement of Consumer Contracts, the Schumer Box, warranty regulation, and the distinction between informational, market, and reputational failures.

**Evidence anchor:** The section traces no-reading rationales through the Restatement debate and disclosure mandates, then argues that growing smart-reader sophistication may shift policy toward uptake while leaving other justifications intact.

**Boundary:** Solving reading does not solve bargaining power, market failure, reputational failure, bias, or unfair terms; the authors therefore do not predict the end of consumer protection or endorse immediate deregulation.

**Connections:** consumer-contract Restatement; meaningful assent; unconscionability; merger clauses; mandatory disclosure; future-proof regulation

**Record:** `ssrn-3740356-p13` · `machine-drafted-source-checked`

## 14. Courts and agencies can use language models to structure corpus-based interpretation and prioritize suspicious contract terms

**Location:** Part IV.B, Courts and Agencies, printed pp. 136-137 (PDF pp. 54-55)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 136–137, that institutions as well as consumers can benefit from smart readers. Courts could operationalize corpus linguistics by using language models to estimate contextual frequencies rather than relying only on introspection or dictionaries, while agencies could scan sector-wide contract sets and flag offensive, suspicious, or unusual terms for human attention. This is significant because it treats automation as a way to allocate scarce judicial and enforcement attention, not only as a retail advice product. It connects to ordinary meaning, corpus linguistics, dictionary use, regulatory triage, supervisory technology, sectoral benchmarking, and human review of algorithmically flagged cases.

**Evidence anchor:** The authors contrast dictionaries with corpus evidence, illustrate probabilistic usage analysis, and propose agency processing of industry contracts to focus limited resources on irregular terms.

**Boundary:** Usage frequencies do not themselves resolve normative meaning, the authors’ numerical example is hypothetical, and agency flags require human investigation because smart readers remain imperfect.

**Connections:** corpus linguistics; ordinary meaning; judicial interpretation; regulatory technology; enforcement triage; contract surveillance

**Record:** `ssrn-3740356-p14` · `machine-drafted-source-checked`

## 15. Existing contract doctrines tend to place innocent smart-reader error on consumers, but a better regime would share incentives through machine-readable disclosure of key terms

**Location:** Part IV.C.1, Allocation of Error Costs, printed pp. 137-140 (PDF pp. 55-58)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 137–140, that ordinary mistake, misrepresentation, misunderstanding, and duty-to-read doctrines offer consumers little recourse when an independent smart reader misstates a contract. Producer liability may improve technology but also raise price and chill entry, while neither buyer nor seller fully controls the risk. They therefore propose adapting conspicuousness, contra proferentem, or the Restatement’s duty to warn so key terms must be disclosed in a smart-reader-friendly form as a condition of enforcement. This is significant because it frames model error as a legal-accident problem requiring incentive-compatible loss allocation rather than automatically blaming the user or developer. It connects to mistake doctrine, misunderstanding, products liability, autonomous-system accidents, conspicuous disclosure, contra proferentem, Restatement section 211, and machine-readable contracting.

**Evidence anchor:** The section tests standard doctrines against several innocent-error hypotheticals, analyzes producer, buyer, and seller incentives, and proposes enforcement-conditioned disclosure of key terms in reader-friendly form.

**Boundary:** The proposal is tentative; seller control over third-party apps is limited, producer liability can suppress beneficial adoption, consumer nonreading complicates cheapest-cost-avoider analysis, and no single party is clearly best positioned to prevent every error.

**Connections:** allocation of error costs; mistake doctrine; products liability; conspicuousness; contra proferentem; machine-readable terms; Restatement section 211

**Record:** `ssrn-3740356-p15` · `machine-drafted-source-checked`

## 16. Courts should not expand the duty to read merely because smart readers appear cheap and accessible

**Location:** Part IV.C.2, The Duty to Read, printed pp. 140-141 (PDF pp. 58-59)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 140–141, that courts and legislatures may prematurely treat smart-reader access as a stronger opportunity to understand contracts and correspondingly expand the duty to read. Repeat players may promote that expectation, fluent demonstrations may obscure technological weakness, and policymakers may try to force adoption through doctrine. This is significant because a legal presumption of technological competence can arrive before reliable tools and equal access do, converting a promised aid into a new burden on consumers. It connects to clickwrap and browsewrap, constructive notice, strategic technology mandates, digital inequality, regressive cross-subsidies, and the institutional lag between technical reality and judicial doctrine.

**Evidence anchor:** The section explains the current duty-to-read rule, identifies repeat-player, judicial-capacity, and strategic-adoption pressures for expansion, and warns that unequal technology access can make a stronger rule regressive.

**Boundary:** Maintaining the existing rule may modestly reduce incentives to adopt useful readers, but the authors regard that cost as smaller than the danger of doctrine outrunning reliability and access.

**Connections:** duty to read; constructive notice; clickwrap; browsewrap; digital divide; premature regulation; consumer inequality

**Record:** `ssrn-3740356-p16` · `machine-drafted-source-checked`

## 17. Because adversarial contract manipulation is hard to detect and prove, legal response will require imperfect combinations of burden shifting, deterrence, and regulatory monitoring

**Location:** Part IV.C.3, The Problem of Adversarial Attacks, printed pp. 141-143 (PDF pp. 59-61)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 141–143, that adversarial attacks resist ordinary enforcement because innocuous-looking choices of spacing, fonts, word order, color, register, or margins can mislead models, while intent is difficult to prove and contract remedies rarely include punitive damages. Statistical evidence that a document misleads a sample of readers might justify shifting the burden to the drafter, and agencies could monitor formatting for suspicious patterns, but both approaches can also penalize innocent drafting for technical model failures. This is significant because neither traditional fraud doctrine nor purely technical detection supplies a comprehensive answer to strategic machine-facing boilerplate. It connects to res ipsa loquitur, burden shifting, optimal penalties for low-detection violations, punitive damages, fraud, CFPB and FTC monitoring, adversarial robustness, and the distribution of false-positive costs.

**Evidence anchor:** The section catalogs possible textual attack surfaces, explains detection and proof obstacles, considers high penalties and fraud, evaluates statistical burden shifting, and recommends ongoing agency and consumer-organization attention.

**Boundary:** Every proposed response is incomplete: attacks may be invisible, intent evidence scarce, penalties constrained, statistical error endemic, and agencies’ own smart readers vulnerable to the same manipulations.

**Connections:** adversarial attacks; burden shifting; punitive damages; fraud; regulatory monitoring; CFPB; FTC; false positives

**Record:** `ssrn-3740356-p17` · `machine-drafted-source-checked`

## 18. Law should prepare for discrimination based on smart-reader use before data-driven personalization becomes entrenched

**Location:** Part IV.C.4, Bias and Discrimination, printed pp. 143-145 (PDF pp. 61-63)

Professors Yonathan A. Arbel and Shmuel I. Becher claim, in “Contracts in the Age of Smart Readers” on pages 143–145, that big-data personalization reverses the old presumption favoring individualized over standardized contracts. If firms offer better terms to smart-reader users and worse terms to nonusers—especially when usage correlates with race or other protected characteristics—the practice can create regressive transfers and eliminate the technology’s positive market spillovers. Unfairness or deception law may reach some instances, but market segmentation doctrine, injury standards, material-omission proof, and benign uses of personalization make a blanket ban undesirable. This is significant because intervention becomes harder after firms collect usage data and build reader status into pricing and contract design. It connects to unfair or deceptive acts and practices, proxy discrimination, algorithmic segmentation, material omissions, freedom of contract, protected classes, precautionary regulation, and path dependence in data markets.

**Evidence anchor:** The final substantive section analyzes discrimination by reader status, tests unfairness and deception theories, rejects a blanket ban, and urges precaution before data collection and tailored treatment become entrenched.

**Boundary:** The legal balance depends on values and future evidence; personalization has legitimate uses, current unfairness and deception theories are contestable, and the authors raise the issue for research rather than prescribe a categorical prohibition.

**Connections:** UDAP law; proxy discrimination; market segmentation; material omission; freedom of contract; precautionary regulation; data-driven contracts

**Record:** `ssrn-3740356-p18` · `machine-drafted-source-checked`