1 of 41

Wikifact

A Project Proposal

2 of 41

Introduction

3 of 41

Synopsis

  • Wikifact will be a statement-centric resource. Each article will be about a statement (and its logical paraphrases).
  • End-users will be able to fact-check arbitrary portions of documents – in Web browsers and word processors – simply by selecting content, using context menus, and making use of services such as Wikifact.
  • End-users will be able to select content from social media websites, news articles, digital textbooks, and arbitrary other documents and websites, and then open context menus providing options for exploring, e.g., fact-checking, the content.

4 of 41

Related Projects

  • AFCNR
  • BRENDA
  • ClaimPortal
  • Full Fact's System
  • Sphere
  • Squash

5 of 41

A Survey

  • In [1], the authors describe interviews with fact-checkers about areas where technology could help: “monitoring potentially harmful content, selecting claims to check, creating and distributing articles, and managing suggestions from readers.”
  • Also discussed were: finding claims worth fact-checking, detecting previously fact-checked claims, evidence retrieval, and automated verification.
  • Some major challenges were noted: multilingual resources, claim ambiguities, system biases, contextual information, and multimodality.

�[1] Nakov, Preslav, David Corney, Maram Hasanain, Firoj Alam, Tamer Elsayed, Alberto Barrón-Cedeño, Paolo Papotti, Shaden Shaar, and Giovanni Da San Martino. “Automated fact-checking for assisting human fact-checkers.” arXiv preprint arXiv:2103.07769 (2021).

6 of 41

Real-time Analytics and Dashboards

  • Wikifact could provide a community dashboard which presents real-time usage data and analytics.
  • Real-time usage data including popular and trending fact-checking articles could be provided so that interested editors could make use of these data to better contribute.

7 of 41

Change Propagation and Notifications

  • Editors could make use of templates to import statements from Wikifact into their Wikinews, Wikipedia, and Wikisource articles.
  • Whenever a statement were updated on Wikifact, those editors expressing interest in dependent articles could be notified so as to be able to revisit those articles and/or to contribute on Wikifact.
  • End-users and editors who visit a new Wikifact article before substantial fact-checking occurs could add those articles to their watchlists, receiving notifications as subsequent fact-checking activities occur.
  • Editors might opt to receive individual alerts, or digests, containing changes or updates relevant to their interests.

8 of 41

Epistemology

9 of 41

Types of Statements

  • Acceptable
    • Does the statement meet the community’s standards, rules and guidelines?
  • Analytic and Synthetic
  • A priori and A posteriori

10 of 41

Annotation: Beyond Boolean Interpretations

  • One can view ascribing Boolean or other truth values to statements as a form of annotation.
  • Annotations utilized by fact-checkers include, but are not limited to: “true”, “mostly true”, “half true”, “partially true”, “mostly false”, “false”, “disputed”, “misleading”, “contains omissions”, “contains exaggerations”, “contains distortions”, “contains hearsay”, and so forth.

11 of 41

Properties of Statements

  • Well-formed
  • Parses, Anaphora resolution, Parts of Speech, Semantic Representations
    • Composition, nested subphrases, e.g., X and Y.
  • Justifications
    • Arguments for and against the statement
    • Corroboration
    • Referenced materials and sources
  • Coherence
  • Relevance
  • Contexts
  • Folksonomic categories
  • Extensibility

12 of 41

Relating Statements

  • Statements can be logical paraphrases of one another.
  • Statements can be also be stylistic variants of one another, where some salient stylistic difference indicates a different intent.
  • Statements can be logically interrelated.
    • Statements can be inferred from sets of statements.
    • Statements can contradict one another.
  • Statements can be related to other things.
    • Wikifact articles could hyperlink to Wikidata entities and Wikipedia articles.
    • Wikifact articles could describe and hyperlink to selections of Web documents containing the statement or a paraphrase of it.

13 of 41

Collections of Statements

  • Statements can be grouped into collections:
    • ones where only one statement in the collection can be logically true, where the statements are mutually exclusive.
    • ones where only the latest statement in the collection is intended to be true, one where the latest statement supersedes previous statements.
    • ones called a “timelines”, “threads”, or “stories”. Whenever a new statement were to be added to such a collection, each previous element could be pinged as a result, notifying interested editors.
    • ones formed as being about the same subject.
    • ad-hoc collections such as those statements co-occurring in documents, articles, reports, or political speeches.

14 of 41

Belief Sets, Models and Theories

  • Wikifact will be inconsistent, though sets of statements can be formed which are each internally consistent.
  • A statement, e.g., a physics-related statement, might be true in a model or theory, e.g., Newtonian or Einsteinian physics.
  • A conceptual model is a representation of a system. It consists of concepts used to help people know, understand, or simulate a subject the model represents.
  • A scientific theory is an explanation of an aspect of the natural world and universe that has been repeatedly tested and corroborated in accordance with the scientific method, using accepted protocols of observation, measurement, and evaluation of results.

15 of 41

Consensus Reality

  • A useful large-scale collection of statements might be one labelled as: “consensus reality.”
    • This could be a set for synthetic, a postiori statements about the world.
      • Statements which describe unfolding events in the world could be added to such a set of statements.
    • While this large set of statements could be inconsistent, informational messages, warnings, and error messages about its consistency can be envisioned as useful to editors.

16 of 41

Types of Justifications

  • Evidence
    • Facts
    • Judgments
    • Testimonies
  • Appeals to authority
  • Arguments from analogy
  • Generalizations
  • Personal experiences

17 of 41

Properties of Justifications

  • Quality rankings would be useful for justifications.
  • With the capability to rank the quality of justifications, e.g., arguments, they could be better displayed for end-users.
    • Alternatively, end-users could upvote and downvote justifications on statements.
      • From this data, at scale, it might be possible to induce algorithms for measuring and ranking justifications in terms of quality.

18 of 41

Frontend and User Experience

19 of 41

Content Selection and Context Menus

  • End-users will be able to fact-check arbitrary portions of documents – in word processors and Web browsers – simply by selecting content, using context menus, and making use of services such as Wikifact.
  • End-users will be able to select content from social media websites, news articles, digital textbooks, and arbitrary other documents and websites, and then open context menus providing options for exploring, e.g., fact-checking, the content.

20 of 41

Paginated Results and Best Guesses

  • When an end-user selects content and makes use of a context menu in a Web browser, resultant content would open in a new tab.
  • This content could provide a paginated list of alternatives or a best guess to a single option with a means of navigating to a paginated list.
    • Real-time usage data and analytics could be of use for enhancing these best guesses which route end-users to specific articles instead of to paginated options.

21 of 41

Incremental Search

  • Another UI/UX concept is that of incremental search, where end-users could enter statements into search textboxes, incrementally, and receive options, as they type, in search suggest drop-down lists.
  • The received options or recommendations would be more elaborated and more contextually specified.

22 of 41

Web Annotations

  • It is possible that, using the W3C Web Annotations data model and protocol, end-users could toggle to visualize selections of content in their documents that have previously been fact-checked or processed on Wikifact.

23 of 41

Backend

24 of 41

Services

  • Context menus on selections of content will utilize services to provide relevant content to end-users, e.g., in new tabs.
  • Wikifact would be such a service.

25 of 41

Wikidata

  • Wikifact could utilize Wikidata as a backend.

26 of 41

Schemas and Ontologies

  • Wikifact would be benefited by formal schemas and ontologies which precisely define the properties of statements, the relationships between statements, and those relationships relating statements and other things.

27 of 41

Use Cases

28 of 41

Journalism

  • Journalism is the production and distribution of reports on the interaction of events, facts, ideas, and people that are both contemporary and relevant, the “news of the day.”
  • Journalism seeks to inform society.

29 of 41

Fact-checking

  • Fact-checking is a process that seeks to verify factual information, in order to promote the veracity and correctness of reporting.

30 of 41

Rational Skepticism

  • Scientific skepticism or rational skepticism, sometimes referred to as skeptical inquiry, is a position in which one questions the veracity of claims lacking empirical evidence.
    • For example, the applicability of hydroxychloroquine for COVID-19.
  • In practice, the term most commonly references the examination of claims and theories that appear to be beyond mainstream science.

31 of 41

Critical Thinking

  • Critical thinking is the analysis of available facts, evidence, observations, and arguments to form a judgement.
  • Several different definitions exist, these generally including the rational, skeptical, and unbiased analysis or evaluation of factual evidence.
  • Critical thinking is self-directed, self-disciplined, self-monitored, and self-corrective thinking. It presupposes assent to rigorous standards of excellence and a mindful command of their use.

32 of 41

Education

  • Education is a purposeful activity directed at achieving certain aims, such as transmitting knowledge or fostering skills and character traits.
  • Various researchers emphasize the role of critical thinking in order to distinguish education from indoctrination.

33 of 41

Science

  • Science is a systematic enterprise that builds and organizes knowledge in the form of testable explanations and predictions about the universe.

34 of 41

Science Communication

  • Science communication is the practice of informing, educating, raising awareness of science-related topics, and increasing the sense of wonder about scientific discoveries and arguments.

35 of 41

Government Transparency & Accountability

  • With the technologies under discussion, people could enhance the transparency and accountability of their governments.

36 of 41

General Public Interest

  • With the technologies under discussion, people could perform real-time fact-checking with respect to arbitrary selections of content from documents in their Web browsers and word processors.

37 of 41

Conclusion

38 of 41

Future Research Directions

  • How might Wikifact interoperate with other AI systems, e.g., Sphere or Project Debater?
  • AI systems, in general, could process and utilize Wikifact data.
  • AI systems, e.g., “bots”, could contribute to Wikifact.
  • Outputs from other AI systems (e.g., answers, arguments, or explanations) could be examined and explored by selecting portions of interest and using context menus to invoke services.

39 of 41

Future Research Directions

  • As selections of documents can contain multimedia, could a Wikifact service scale to include fact-checking and discussing pictures?
  • Earlier, real-time analytics and dashboards were mentioned as being useful to editors. Might AI systems be additionally useful, in these regards, asking editors questions as statements are added or recommending to editors other statements to fact-check?

40 of 41

What Do We Need?

  • More project champions, co-proposers.
  • More project supporters, those expressing interest on the project proposal page: https://meta.wikimedia.org/wiki/Wikifact .
  • More research and development.
  • Please spread the word about the project proposal!

41 of 41

Thank you

  • Questions, comments, feedback?
    • Adam Sobieski (adamsobieski@hotmail.com)