University of Sussex
Browse

File(s) under permanent embargo

Towards collaborative Human-AI systems for knowledge acquisition, modelling, and discovery across domain boundaries

presentation
posted on 2023-06-10, 04:24 authored by Ronald GrauRonald Grau
Collaborative Human-AI systems can help humans address hard problems that are currently difficult to investigate at the system level. Creating software that facilitates human knowledge specification through effective visual interfaces can be augmented with computational techniques to support the cognitive operations that are required for knowledge acquisition, modelling, and discovery of heterogeneous knowledge. For example, there are process systems that exhibit simultaneous physical, chemical, and biological changes at different physical and temporal scales. These systems exist both in nature (e.g., climate science) and in industry (e.g., food production), with interactions that are complex and dynamic in many ways, spanning multiple areas of science. The related knowledge tends to be incomplete, exists at different levels of granularity and abstraction, and is expressed using various concepts, notations and paradigms that may be difficult to represent and reason over in combination. This paper is about a conceptual framework that comprises a representational system, meta model, as well as computational and cognitive methods aimed at assisting humans in the construction of knowledge models, and applied in a food processing domain. Following a representational epistemic approach, interactive diagrams were designed and implemented within a software suite that support cognitive operations for compositional knowledge specification. The software uses various techniques to augment human cognitive abilities, for instance by propagating and making visible changes in the system, across different levels of abstraction, covering multiple perspectives (e.g., compositional structure, changing properties, functional impact). Algorithms and visual cues are also used for guiding human knowledge operations, e.g., by generating new interaction contexts automatically, based on modelling decisions. Key to the success of the system has been how it interfaces with humans at the knowledge level to provide effective means for them to inspect and interact with the model.

History

Publication status

  • Published

File Version

  • Accepted version

Presentation Type

  • paper

Event name

1st International Workshop on Knowledge Representation for Hybrid Intelligence

Event location

Amsterdam, Netherlands

Event type

workshop

Event date

14 Jun 2022

Department affiliated with

  • Informatics Publications

Research groups affiliated with

  • Creative Technology Publications

Full text available

  • No

Peer reviewed?

  • Yes

Legacy Posted Date

2022-08-04

First Compliant Deposit (FCD) Date

2022-08-04

Usage metrics

    University of Sussex (Publications)

    Categories

    No categories selected

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC