Towards Improved Transparency through Computational Terms of Service Agreements

tags

Bogdana Rakova Megan Ma

Notes

computational ToS agreements and computable verification

NOTER_PAGE: (1 . 0.3465909090909091)

“I agree to the terms of service” is perhaps the most falsely given form of consent

NOTER_PAGE: (1 . 0.35714285714285715)

embedding directly into ToS agreements a contractual mechanism to report algorithmic harm

NOTER_PAGE: (1 . 0.38961038961038963)

reports done through the proposed mecha- nism are anonymized, publicly available, and explainable

NOTER_PAGE: (1 . 0.43993506493506496)

empower individuals by offer- ing better literacy of legal agreements

NOTER_PAGE: (1 . 0.6728896103896105)

users are typically unaware of these projects

NOTER_PAGE: (1 . 0.7175324675324676)

ToS and EULAs, traditionally, do not intend to enable users to seek recourse

NOTER_PAGE: (1 . 0.7711038961038962)

leaving users without any means for appeal

NOTER_PAGE: (1 . 0.8214285714285715)

define [algorithmic] harms as incidents experienced by individuals and communities that lead to sociomaterial harms, resulting from algorithmic systems and interactions between human and algorithmic ac- tors

NOTER_PAGE: (2 . 0.12581168831168832)

ensuring that the evaluative construct of ‘impact’ remains close to the sociomaterial harms

NOTER_PAGE: (2 . 0.38961038961038963)

ToS provisions are ambiguous, in- consistent, and lack context.

NOTER_PAGE: (3 . 0.13068181818181818)

three broad kinds of downstream harms of algorithmic systems including privacy violations, unfair competition, and behav- ioral manipulation

NOTER_PAGE: (3 . 0.17451298701298704)

three ways in which sophisticated algo- rithms can generate systemic harm: by facilitating proxy discrimination, by enabling surveillance capitalism, and by inhibiting competition in markets

NOTER_PAGE: (3 . 0.23782467532467533)

Diver [22] brings forth the concept of computational legalism

NOTER_PAGE: (4 . 0.25)

In effect, Diver’s notion of delay, offered by the affordance of text, is akin to the notion of friction

NOTER_PAGE: (4 . 0.4918831168831169)

some perceivable “inefficiencies” are worth maintaining.

NOTER_PAGE: (4 . 0.523538961038961)

divorces from its meaning the essential cognitive exercise (“the meeting of the minds”) from the physical act (“the handshake”)

NOTER_PAGE: (4 . 0.6866883116883117)

ToS agreements should be framed in a manner that makes transparent the normative demands contained in the content of the agreement

NOTER_PAGE: (4 . 0.8530844155844156)

Eigen illustrated that individuals who were provided explicit direction around the terms of the survey agreement, even if inconsequential, were more likely to follow through with contractual obligations.

NOTER_PAGE: (5 . 0.7443181818181819)

merely the act of making contractual terms understandable has the potential to empower individuals.

NOTER_PAGE: (5 . 0.8433441558441559)

the use of logic programming suggests that law, to an extent, is built on logic. It carries forth theoretical traditions of legal positivism

NOTER_PAGE: (6 . 0.14285714285714288)

Computational law seeks to unpack the ontological pro- cess latent to legal text construction.

NOTER_PAGE: (6 . 0.2491883116883117)

computational law has, at its core, clarification, empowering individuals to make sense of legal information that is relevant to them.

NOTER_PAGE: (6 . 0.5073051948051949)

algo- rithmic harms and non-algorithmic harms driven by user behavior, cannot be fully separated in practice.

NOTER_PAGE: (6 . 0.8262987012987013)

embed a mechanism in the computational ToS for users to report experiences of algorithmic harms. Submitted reports are then evaluated through a verification script written in a logic programming language.

NOTER_PAGE: (7 . 0.1103896103896104)

We envision that the reporting mechanism, embedded in the computational ToS, could be implemented through a browser extension. In this way, it would run within users’ browsers, and therefore, would not require platform com- panies to adjust their interfaces.

NOTER_PAGE: (7 . 0.12987012987012989)

free text explanation articulating the circumstances of the perceived harm.

NOTER_PAGE: (7 . 0.21996753246753248)

users can choose to add their incident to a database hosted by an independent civil society organization

NOTER_PAGE: (7 . 0.4626623376623377)

YouTube’s response was limited to removing the videos from the channel. It is unclear if and how their recommender system was changed.

NOTER_PAGE: (8 . 0.8368506493506495)

YouTube requires fundamen- tally an understanding of the Contract, in its entirety, from its users. Contract literacy is a prerequisite to contractual assent.

NOTER_PAGE: (9 . 0.3092532467532468)

modal verbs may be contrasted between these two clauses: from obligation to possibility; the heaviness and inflexibility of the users’ responsibilities against the consideration and permissibility of YouTube’s responsibilities.

NOTER_PAGE: (9 . 0.34983766233766234)

ToS belongs to a network of documents that, in aggregate, constitute the contract

NOTER_PAGE: (9 . 0.45292207792207795)

This could be incredibly onerous for the user; not only to determine which category the harm experienced would fall under, but also how to then assess the ways in which the content had violated the ToS. On the other hand, if the conditions governing each particular category are embedded as code, the type of harm experienced could be verified immediately and transparently. Users need only to identify the characteristics of the perceivably inappro- priate content for the reporting mechanism to describe the category of harm and the reason it violates the ToS.

NOTER_PAGE: (10 . 0.1866883116883117)