Principles of High Quality Documentation for Provenance: A Philosophical Discussion

Abstract

Abstract.Computer technology enables the creation of detailed documentation about the processes that create or affect entities (data, objects, etc.). Such documentation of the past can be used to answer various kinds of questions regarding the processes that led to the creation or modification of a particular entity. The answer to such questions are known as an entity's provenance. In this paper, we derive a number of principles for documenting the past, grounded in work from philosophy and history, which allow for provenance questions to be answered within a computational context. These principles lead us to argue that an interaction-based model is particularly suited for representing high quality documentation of the past.History is important: in order to make progress in the future, it is important to learn the lessons of the past. History, therefore, becomes a vital resource for progressive societies. History is based on evidence or documentation of events that occurred in the past. Computer technology enables documentation to be produced that is more accurate and comprehensive than previously possible. Given the quantity of documentation that can be generated by computer systems (for example those running e-Science applications), what kind of documentation should be created that enables historians and users to most effectively answer questions they have about the past. To answer these questions, we argue for two principles for the creation of, what we term, high quality documentation of the past. One principle guarantees that users of documentation of the past have a precise semantics for it. The other principle guarantees that there exists a link between documentation and its creators. Together these principles are the first contribution of this paper.A question that is often posed by historians and users regards the provenance of an entity (object, data item, etc..): what was the process that led to the entity in question? To enable the answering of provenance questions in a manner that conforms with the aforementioned principles, we introduce a model of documentation based on the exchange of messages between actors. The justification of this model is the second contribution of this paper.The rest of this paper is organised as follows. We begin with a more detailed motivation of our work. After which, three assumptions about the world are

Toward a Theory of Self-explaining Computation

“…[32,37,39,31]). This work, like a great deal of work on provenance, has invoked informal motivations such as that provenance should identify data that are "relevant to", "caused" or "influenced" an output and provide "repeatability", "transparency", or "explanation".…”

A Provenance Model for Manually Curated Data

“…This is a common situation, and we believe our approach meets most of the criteria above; some issues, such as robustness in the face of changes to other databases, are not yet handled. As remarked by Groth et al [10], it is important to think carefully about what provenance information to retain, based on what information will be useful evidence to later observers. In our setting, we aim only to track the internal relationships among a sequence of versions of the target database and fixed source databases.…”

Electronically Querying for the Provenance of Entities

“…The separation allows us to tailor systems to best meet the requirements on each. For example, the details of a process may not be recoverable after it has completed, so should be documented as completely and accurately as possible during or immediately after [7], while provenance can be determined as long as process documentation is preserved unmodified, so should be scoped to include only that relevant to the querier.…”

“…A workflow, shown in Figure 1, is run and process documentation generated: a GUI actor calls an Averager service with two values (7,5); Averager sums them and calls Divider with 12 as divisor, 2 as dividend; Divider sends the answer 6 to Averager, which sends it to the GUI; the GUI sends 6 to Store, along with the file location, e.g. file1, at which to store it.…”

“…Store Average (7,5) Answer (6) Answer (6) Divide (12, 2) Store (6, file1) Entity for which we want to find the provenance…”