Leitura de Artigo - Foco no Exemplo de Regras

Knowledge Graphs: Trust, Privacy, and Transparency from a Legal Governance Approach

DOI: https://doi.org/10.26826/law-in-context.v37i1.126

 

*** Daniel é o principal autor ***

Although to this day there isn’t a precise definition of the term (Ehrlinger and Wöß 2016, Hogan et al, 2020), we adopt the view that a Knowledge Graph (KG) represents a network of interlinked descriptions of entities (objects, events, concepts etc.)-- a graph-theoretic representation of human knowledge such that it can be ingested with semantics by a machine (Kejriwal 2019).

 

[A definição de KG] 

While the graph model or some variant has been used in several KGs, it has already been observed that using only atomic (indivisible) nodes as the “granule” of information is insufficient to express complex types of information, such as events, or time-varying data. This highlights the fact that data ultimately expresses a belief, opinion or point of view of some agent – the author.

 

[Contexto temporal]

[Considerar como verdade depende do contexto da informação e do contexto de uso]

 

Fourth Industrial Revolution (4IR) (Schwab 2017). The 4IR is characterized by a fusion of technologies, which is blurring the lines between the physical, digital, and biological spheres.

 

As we will explain later, regulations do not comprehend only legislation, i.e. enacted laws and statutes, but all kinds of legal instruments of governance—hard law, soft law, policies and ethics.

 

In order to be effective, we claim that the use of Knowledge Graphs ought to provide support for these concerns—trust, privacy and transparency. Accordingly, privacy and data protection can be considered from the legal governance approach. This means that they can be implemented through the construction, development and implementation of a technological toolkit, comprising data mining, data analytics and the linked open data tools of the later developments of the Semantic Web—i.e. the Web of Data. The Knowledge Graphs approach that we embrace in this article is related to this dimension.

 

However, new legal issues arise, such the use by LOD of crowdsourced vocabularies, where there is no authority imposing one interpretation over another. There is no evidence so far of case-law nor out-of-court disputes regarding linked data resources. Legal resources should be differentiated from legal sources. The former refers to the large number of existing legal vocabularies and documents on the Web of Data. The latter refers to the specific content that ‘counts as legal’ at regional, national, international and transnational levels to be effectively implemented or enforced. Determining what is ‘valid’ law, what counts as legal, is in itself a non-trivial theoretical operation that is usually performed through the concepts of doctrine, legal theory, and checks-and-balances

 

3. BACKGROUND CONCEPTS

3.1 TRUST

 

The original vision for the Semantic Web included a “Trust” layer, although its emphasis was more on authentication and validation with static trust measures for data. There have been many efforts in representing trust, including computational models.

 

[Significado para Semantic Web]

 

... that trust is “knowledge-based reliance on received information”, that is, an agent (i.e., a person or a software program) decides to trust (or not) based solely on her/his knowledge, and the decision to trust implies the decision to rely on the truth of received or on already known information to perform some action.In terms of a Knowledge Graph, an agent wishing to perform an action must first filter those information items it deems “trusted”, i.e., it will use them to perform the intended action. Since it is not possible to “half-act”, in this sense trust is binary—either the agent trusts the information, or it does not.

 

[Definição de Confiança para consumo de informação dos KGs]

 

3.2 PRIVACY

 

Based on this theory, we define Privacy as “controlled access to information related to an agent”. In order to ensure privacy, it is necessary to answer three questions:
• Q1: What types of Actions are allowed (and controlled) over Knowledge Items (KIs)?
• Q2: What are the relation types between some Agent and a KI that entitle this Agent to establish a Privacy Rule governing Actions over that KI?
• Q3: How to resolve conflicts between applicable rules?

 

[Definição de Privacidade]

 

3.3 TRANSPARENCY

 

From a more abstract approach, both Privacy and Transparency relate to controlling actions over information, and who can define such controls. As such, they can be regarded as two points in the same control dimension. Privacy tends to limit or restrict actions over information items, whereas Transparency tends to allow (in some cases, mandate) actions over them, which explains the natural tension that exists between the two.

 

4. AN ILLUSTRATION - DISASTER RELIEF DONATION

 

5. A SUMMARY OF THE KG USAGE FRAMEWORK.

 
Figure 3 shows a diagram of the use of information within a KG. “Using a KG” is represented as a Request made by some Agent for an Action over a Knowledge Item (KI). 

 

Thus, a Knowledge Graph represents a collection of interlinked descriptions of KIs – real-world objects, events, situations or abstract concepts.We propose to represent the KG as a collection of Knowledge Items (KIs), each of which as a nanopublication (Groth, Gibson and Velterop 2010). A nanopublication “offers a supplementary form of publishing alongside traditional narrative publications”, consisting of three parts representable by RDF graphs: “(i) an assertion (a small, unambiguous unit of information), (ii) the provenance of that assertion (who made that assertion, where, when, etc.), (iii) the provenance of the nanopublication itself (who formed or extracted the assertion, when, and by what method).”

 

[Afirmação contextualizadas por duas camadas de proveniência]

 

5.1 KNOWLEDGE ITEM REPRESENTATION

 
A KI, as all nanopublications, comprises an assertion graph, a provenance graph and a publication info graph. The assertion graph of a KI contains a set of assertions about its content. The assertions in this graph are a subset of the assertions in the underlying RDF graph.

 

The provenance graph of the nanopublication will contain provenance information about the assertions in the assertion graph (e.g.; what image or natural language processing software was used, recorded location info, whether the assertions were inferred using some inference engine, etc.). The provenance graph can be used to represent, to the desired level of detail, the supporting information for the assertions. 

Another use of provenance can be seen in the case of a statement stating that, for example, <Barack Obama> placeOfBirth <Hawaii>. The provenance information may include documentation to support its truthfulness, such as a reference to a birth certificate that states that indeed the place of birth of Barack Obama is Hawaii. The publication info graph will contain metadata about.

 

[Exemplo de Afirmação que tem disputa na Wikidata]

 

5.2 CONTROLLING USAGE

 

In our framework, Privacy and Transparency refer to Request for actions over some information (in a KI), for which an Authorization must be granted, according to the Rules set forth by the relevant stakeholders. Stakeholders include persons “related” to the KI, as well as institutional agents such as “the State” (whose rules are stated as laws). Rules may be based (make use of) on any information available in the KG. We refer to both Privacy and Transparency rules collectively as Usage Rules.

...

 

5.3 RULES

 

Rules are of the form antecedent => consequent, both of which are sets of statements (Almendra and Schwabe 2016).

 

6. EXAMPLE SCENARIO REVISITED

 

The trust rule below captures this, expressed using N3Logic with extensions. 

 

We state under KG some statements we assume to be present in the KG:
KG
NonProfit subClassOf org:Organization. 

AuditCo subClassOf org:Organization.
CertificationAgency subClassOf org:Organization.
Donate subClassOf Action.
<ReliefOrg> type NonProfit. 

<AuditInc> type CertificationAgency.

<ReliefOrg> officer <George>. 

<AuditInc> officer <George>.

 

RuleEd1
[SE]{

?O type org:Organization; hasOfficial: ?Ofc; hasFinancialRecord ?FR. 

?Ofc type foaf:Person.
?FR assertions ?FRa. 

?FRa log:semantics ?FRaS.

?FRaS log:includes { ?FR auditedBy ?Aud. ?Aud type AuditCo}.
?FR provenance ?PFr. 

?PFr log:semantics ?PFrS. 

?PFrS log:includes {prov:hasPrimarySouce ?DOCS}. 

?DOCS assertions ?DOCSa. 

?DOCSa log:semantics ?DOCSaS.
?DOCSaS log:includes {AuditCo certifiedBy ?CA. }.
<TrustedGraphEd> author <Ed>; log:semantics ?TGEd. 

?TGEd log:includes {?CA type CertificationAgency}

}

=> 

[ENTÃO]{<TrustedGraphEd> :add {?O type:NonProfit}. }

               

[Conjuntiva]

 

RuleEd2
{

<TrustedGraphEd> author <Ed>; log:semantics ?TGEd.
?TGEd log:includes {?O type:NonProfit, officer:<George>}. <ruleEd1> author <Ed>}.
ruleEd1 assertions ?ARule1. 

?Arule1 log:semantics ?Arule1S.

?ARule1S log:includes {{ <act1> type Donate, recipient <?O>. <Ed> intends <act1>}

}
=> 

{<at> type Authorization, rule <RuleEd2>, action <act1>, value “Denied”}}

 

RuleGeorge1
{G :is {?O type:NonProfit; officer:<George>. 

<TrustedGraphGeorge> author <George>; log:semantics ?TGG.
?TGG log:includes {?O officer:<George>. <ruleGeorge1> author <George>}.
ruleGeorge1 assertions ?AGeorge1. 

?AGeorge1 log:semantics ?AGeorge1S. 

?Ageorge1S log:includes { <act1> type Read; object ?G. ?A intends <act1>}
=> 

{<at> type Authorization; rule <RuleGeorge1>; action <act1>; value “Denied”}}

[Regra SE/ENTÃO ... CONSTRUCT {ENTÃO} WHERE {SE} ]

7. CONCLUSIONS

 

We have shown how legal requirements, and other types of norms, which ultimately regulate the functioning of any application that uses the KG, can be incorporated into the KG itself.

Almendra, V. D. S., and Schwabe, D. 2006. Trust policies for semantic web repositories. In Proceedings of 2nd International Semantic Web Policy Workshop (SWPW’06), at the 5th International Semantic Web Conference, ISWC, pp. 17-31

 

Artz, D., Gil, Y. 2007. “A survey of trust in computer science and the Semantic Web”, Web Semantics: Science, Services and Agents on the World Wide Web, 5 82): 58-71. http://www.sciencedirect.com/science/article/pii/S1570826807000133

Zuboff, S., 2019. The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. Barack Obama's Books of 2019. NY: Profile Books.

 

================================================================

 

N3Logic: A Logical Framework For the World Wide Web

N3Logic uses N3 syntax and extends RDF with a vocabulary of predicates. N3 aims to do for logical information what RDF does for data: to provide a common data model and a common syntax, so that extensions of the language are made simply by defining new terms in an ontology. 

 

Another goal of N3Logic is that information, such as but not limited to rules, which requires greater expressive power than the RDF graph, should be sharable in the same way as RDF can be shared. This means that one person should be able to express knowledge in N3Logic for a certain purpose, and later independently someone else can reuse that knowledge for a different unforeseen purpose. As the context of the latter use is unknown, this prevents us from making implicit closed assumptions about the total set of knowledge in the system as a whole.

 

We’ve chosen to adopt a monotonicity requirement for N3Logic because we find it scales well. This implies that the addition of new information from elsewhere cannot silently change the meaning of the original knowledge, though it might cause an inconsistency by contradicting the old information. 

 

Certain syntaxes (e.g. “[“...”]”) allow an existential variable to be introduced without having to name it: this is known as a blank node. The following are equivalent:
 

@forSome X. j:Joe foaf:knows X. X foaf:name "Fred" .
j:Joe foaf:knows [ foaf:name "Fred" ] .

 

N3 extends the abstract syntax of RDF in two ways: 

It has all of the terms of RDF plus quoted formulas. 

It has all of the formulas of RDF plus universally quantified formulas. In simple cases, the @forAll quantifier can be left implicit.

 

4.3 N3Logic Vocabulary

 
N3Logic uses the N3 syntax and also includes a set of predicates. Its vocabulary is union of the N3 syntax and the set of URI references defined in the log: (http://www.w3.org/2000/10/swap/log#), crypto: (http://www.w3.org/2000/10/swap/crypto#), list: (http://www.w3.org/2000/10/swap/list#), math: (http://www.w3.org/2000/10/swap/math#), os: (http://www.w3.org/2000/10/swap/os#), string: (http://www.w3.org/2000/10/swap/string#), and time: (http://www.w3.org/2000/10/swap/time#) namespaces as shown in Table 4.3.

 
log:conclusion, log:content, log:includes, log:semantics, log:notIncludes, log:supports ... crypto:md5, crypto:sign, crypto:verify ... list:in, list:last ... math:lessThan, math:greaterThan ... os:argv, os:environ ... string:contains, string:endsWith, string:scrape ... time:day, time:hour, time:minute ... 

 
Table 1. Some N3Logic predicates