What is the notion of trust, when it comes to publishing linked open data in the cultural heritage sector? This presentation discusses some aspects with relation to three primary questions: How do we trust what was said, trust that the institution said it, and trust what it means?
5. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
So … Linked Open Data?
• The Web of Data
• Institutions publish data on the web, just like web pages
• At URLs under their domain name
• The data uses shared standards
• We’ll come back to these at the end
• The data has references to other bits of data
• … Published by the same institution
• … Published by other institutions
8. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
“Trust” in the Data?
• Accuracy: Does the data correctly represent the state of the
real world for the things it describes? (Objective)
• Certainty: Belief of the Publisher as to the extent of the
accuracy of the data. (Subjective)
• Utility: Belief of the Researcher that the data is useful for
fulfilling their current information need. (Subjective, time-
specific)
13. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Relations between Actors
• Confidence: Belief of the consumer in the current and past
competence of the publisher (accuracy of data)
• Trust: Belief of the consumer in the current and future
benevolence of the publisher (availability of accurate data)
• Dependence: When one actor relies on another for the
successful outcome of a critical function
18. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Personal or Institutional Trust?
• Trust: From the reputation of the Institution
including by considering confidence over time
• Data: Created by actions & expressing the beliefs of
many Individuals over time
Linked Open Data is relatively new
Our trust in the institution likely predates the LOD
Should our trust be in the People,
not the Institution?
22. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Trust: Similarity
People tend to view the actions of other members of groups
with which they identify (in-groups) more favorably than
those of members of groups with which they do not.
Even when the action is identical.
We are more likely to trust organizations that seem to share
similar missions, constituencies or worldviews to our own.
Summary: https://en.wikipedia.org/wiki/In-group_and_out-group
24. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Structuring Data is a Political Act
with Ethical Implications
https://lincsproject.ca/events/lincs-conference-2021/
• Information systems enact systems of power
• All data are contextual, discursively produced, and
historically contingent
• There are many different kinds of knowledge, and ways to
come to know things
-- Erin Canning, LINCS Project
(Linked Infrastructure for Networked Cultural Scholarship)
34. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Design Patterns for Usability
1. Scope design through shared use cases
2. Design for international use
3. As simple as possible, but no simpler
4. Make easy things easy, complex things possible
5. Avoid dependency on specific technologies
6. Use REST / Don’t break the web
7. Separate concerns, keep APIs loosely coupled
8. Design for JSON-LD, using LOD principles
9. Follow existing standards & best practices, when possible
10. Define success, not failure (for extensibility)
https://iiif.io/api/annex/notes/design_patterns/, https://linked.art/api/1.0/principles/
40. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Web Archiving
Public archiving of open data:
• Trusted third party to ensure integrity and availability
• Many CH organizations already engaged with web
archiving, for digital preservation purposes
• Allows “time travel” through different versions
• Across multiple archives
• Memento / IETF RFC 7089
http://mementoweb.org/guide/, esp. http://dx.doi.org/10.1109/MIC.2012.78
47. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Standards Need Same Considerations as
Data
If the standard conveys the meaning, how do we trust the
standard?
• Sustainable products with governance and institutions
• Archived, versioned and harvestable
• Usable by data modelers and developers
• Useful to content specialists and researchers
• Diversity of institutions and people, in every role
48. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
• Conceptual Model
• Abstract way to think about the world,
holistically, consistently and coherently
• Ontology
• Shared set of terms to encode that thinking
in a logical, machine-actionable way
• Vocabulary
• Curated set of sub-domain specific terms,
to make the ontology more concrete
encodes
refines
Model
Ontology
Vocabulary
Abstraction Standards
50. Trust
and
Belief
in
Linked
Open
Data
robert.
sanderson
@yale.edu
@azaroth42
Example: Linked Art
• Model: CIDOC Conceptual Reference Model
• Ontology: RDF encoding of CRM 7.1, plus extensions
• Vocabulary: Getty AAT, plus minimal extensions
• Profile: Art Museum oriented
(but dips toes in adjacent domains)
• API: JSON-LD format with
10 primary divisions of content
See: https://linked.art/
What do you need to know about Linked Data for this presentation to make sense? Don’t worry, there’s no need to talk about triples, inference, RDF, or anything like that.
Conflation of three separable features.
Looking for hair in materials for an analysis of human remains in a collection, then this record is very useful – high utility for that research question, but low for most others given the uncertain (but perhaps accurate) information.
Looking for the oldest person in our data… 39 trillion years old.
When we talk about trust, we often mean confidence.
Accuracy is needed, of course. Need to be confident that the institution knows what it is talking about.
People come and go, but we want the data to remain open and available. Needs to have the right governance and motivations to continue to spend its resources for the good of the community
Another relevant factor for trust is similarity between trustor and trustee. I would trust Getty over Google, even if they both published the same data.
Resist the centralization and homogenization of knowledge, for that is the new cultural colonization. We need diverse ecosystems of data, not a single organization or monoculture.
But don’t worry, it’s about to get worse when we add in the next actor…
So the researcher’s belief about the utility of the data, is dependent on the application’s rendition of that data. The same data rendered to the user in a different way might evoke different belief in its utility.
Utility of the data in aggregate across many researchers and users is a large factor in external impact. And external impact is a factor in our trust in the institution.Consider Wikidata – Usability, mostly from having a lot of information in one spot, and assuming some basic level of accuracy, building on the confidence of a similar but unrelated product – Wikipedia.
3 & 4. Balance of Accuracy vs Usability – encoding the structured expression of certainty in the data reduces usability at the expense of marginal benefit over expressing it only in a human-understandable form
9 – Coming back to standards in few.
30 MINUTES
But before standards…
Adam Soroka of the Smithsonian
Yes, you in the back who looks like the know the answer… yeah, no.
But if you do have a spare 69.3 million dollars, I also have some linked metadata about images to sell you.
Reducing the need for trust in the institution, only confidence. Can archive locally as well to avoid any third party trust.
35 MINUTES
Now we have a bunch of lego bricks
The profile tells us how to put those lego bricks together in a useful way.
The API tells the technologists how to get access to the results in a usable way.