Interdisciplinary Symposium on Cognitive Ontologies (ISCO)
Date and Time
Date and Time
The symposium will take place on the 26th of May, from 2:15 pm - 9:15 pm (CEST).
The symposium will take place on the 26th of May, from 2:15 pm - 9:15 pm (CEST).
Description
Description
In the ISCO, experts from diverse fields of research come together to discuss current advances in the development of cognitive ontologies to aid knowledge accumulation and integration in Cognitive Neuroscience. We will discuss the challenges and chances of ontology-based knowledge representation and management by addressing issues such as current thoughts and practices of conceptualising cognitive functions, top-down and data-driven approaches in developing cognitive ontologies, the use of formal ontologies for (FAIR) knowledge engineering, controlled vocabularies for automated data retrieval and reasoning as well as future perspectives on developing a comprehensive cognitive ontology.
In the ISCO, experts from diverse fields of research come together to discuss current advances in the development of cognitive ontologies to aid knowledge accumulation and integration in Cognitive Neuroscience. We will discuss the challenges and chances of ontology-based knowledge representation and management by addressing issues such as current thoughts and practices of conceptualising cognitive functions, top-down and data-driven approaches in developing cognitive ontologies, the use of formal ontologies for (FAIR) knowledge engineering, controlled vocabularies for automated data retrieval and reasoning as well as future perspectives on developing a comprehensive cognitive ontology.
Invited Speakers
Schedule
14:15-14:20 OPENING
14:15-14:20 OPENING
14:20-14:25 SESSION 1: CONCEPTUALISING COGNITIVE FUNCTIONS (INTRO)
14:20-14:25 SESSION 1: CONCEPTUALISING COGNITIVE FUNCTIONS (INTRO)
14:25-14:45 A. M. JACOBS: WHEN WILL WE EVER LEARN? MUSING ABOUT THE STATE OF THE ART IN MODELING BRAIN AND BEHAVIOR PROCESSES.
14:25-14:45 A. M. JACOBS: WHEN WILL WE EVER LEARN? MUSING ABOUT THE STATE OF THE ART IN MODELING BRAIN AND BEHAVIOR PROCESSES.
14:45-14:55 INTERIM DISCUSSION
14:45-14:55 INTERIM DISCUSSION
14:55-15:00 SESSION 2: STRUCTURE-FUNCTION MAPPING (INTRO)
15:00-15:20 R. POLDRACK: WHY DO WE NEED A FORMAL ONTOLOGY OF COGNITION, AND WHAT SHOULD IT LOOK LIKE?
15:20-15:25 SHORT QUESTIONS
15:25-15:45 J. GOMEZ-LAVIN: MOSAIC ONTOLOGIES OF COGNITION: A TEST CASE FROM WORKING MEMORY.
15:45-15:50 SHORT QUESTIONS
15:50-16:05 INTERIM DISCUSSION
16:05-16:35 BREAK
16:35-16:40 SESSION 3: MACHINE-ACTIONABLE CONTROLLED VOCABULARIES (INTRO)
16:40-17:00 K. ROBBINS: HED - MOVING BEYOND KEYWORDS TO FINE-GRAINED ANALYSIS-READY ANNOTATION IN NEUROIMAGING.
17:00-17:05 SHORT QUESTIONS
17:05-17:25 J. TURNER: BRIDGING THE GAP FROM PUBLICATION TO PRACTICE IN NEUROIMAGING DATA SHARING.
17:25-17:30 SHORT QUESTIONS
17:30-17:45 INTERIM DISCUSSION
17:45-17:55 BIO BREAK
17:55-18:00 SESSION 4: FORMAL ONTOLOGIES AND REPRESENTATION LANGUAGES (INTRO)
18:00-18:20 G. GUIZZARDI: OBJECT META-TYPES, CROSS-WORLD IDENTITY AND CONCEPTUAL SPACES.
18:20-18:25 SHORT QUESTIONS
18:25-18:45 B. SMITH: FROM BFO TO MFO: A MODULAR APPROACH TO COGNITIVE ONTOLOGIES.
18:45-18:50 SHORT QUESTIONS
18:50-19:10 INTERIM DISCUSSION
19:10-19:55 BREAK
19:55-20:00 SESSION 5: FUTURE PERSPECTIVES ON COGNITIVE ONTOLOGIES (INTRO)
20:00-20:10 A. RAVENSCHLAG: CAN A MULTI-MODULAR ONTOLOGY AID PROGRESS IN COGNITIVE NEUROSCIENCE?
20:10-20:20 M. DENISSEN: A GATEWAY TO DATA: EXPERIMENTS AS A POTENTIAL HUB BETWEEN CONCEPTUAL KNOWLEDGE AND NEUROIMAGING DATA.
20:20-20:30 INTERIM DISCUSSION
20:30-21:00 PANEL DISCUSSION
21:00-21:15 CLOSING REMARKS
Session Overview
Session Overview
SESSION 1 introduces commonly used terminologies and practices in Cognitive Neuroscience with emphasis on the necessity for a general framework for evaluating computational models and the significance of defining cognitive concepts based on such models. We will discuss the value of cognitive theories as organizing principles for neurocognitive research as well as practical implications of integrating diverging theoretical perspectives.
SESSION 1 introduces commonly used terminologies and practices in Cognitive Neuroscience with emphasis on the necessity for a general framework for evaluating computational models and the significance of defining cognitive concepts based on such models. We will discuss the value of cognitive theories as organizing principles for neurocognitive research as well as practical implications of integrating diverging theoretical perspectives.
SESSION 2 introduces current advances in ontology development for the domain of Cognitive Neuroscience. We will discuss the benefits of top-down and data-driven approaches when developing cognitive ontologies, as well as alternatives to contemporary definitions of cognitive concepts as natural kinds and potential ways to assign these to an ontological architecture.
SESSION 2 introduces current advances in ontology development for the domain of Cognitive Neuroscience. We will discuss the benefits of top-down and data-driven approaches when developing cognitive ontologies, as well as alternatives to contemporary definitions of cognitive concepts as natural kinds and potential ways to assign these to an ontological architecture.
SESSION 3 introduces controlled vocabularies for machine-actionable representations of neuroimaging experiments and data. We will discuss the requirements for controlled experimental classification and the description of complex datasets as well as machine-actionable representations of neuroimaging paradigms and the potential of cognitive ontologies for automated data retrieval and open data sharing.
SESSION 3 introduces controlled vocabularies for machine-actionable representations of neuroimaging experiments and data. We will discuss the requirements for controlled experimental classification and the description of complex datasets as well as machine-actionable representations of neuroimaging paradigms and the potential of cognitive ontologies for automated data retrieval and open data sharing.
SESSION 4 addresses the role of formal ontologies and representation languages for (FAIR) ontology engineering. We will discuss the value of top-level ontologies for data integration and the means of modern information systems engineering, as well as the benefits of cognitive ontologies as interfaces for complex data query and the meaning of ontology implementation in the context of Open Science.
SESSION 4 addresses the role of formal ontologies and representation languages for (FAIR) ontology engineering. We will discuss the value of top-level ontologies for data integration and the means of modern information systems engineering, as well as the benefits of cognitive ontologies as interfaces for complex data query and the meaning of ontology implementation in the context of Open Science.
SESSION 5 addresses future perspectives on ontology development, including potential benefits of conceptualising a multi-modular ontology for knowledge representation and data management. We will discuss domain-specific requirements for ontology development based on cognitive concepts and neuroimaging data as well as potential functionalities of a comprehensive cognitive ontology.
SESSION 5 addresses future perspectives on ontology development, including potential benefits of conceptualising a multi-modular ontology for knowledge representation and data management. We will discuss domain-specific requirements for ontology development based on cognitive concepts and neuroimaging data as well as potential functionalities of a comprehensive cognitive ontology.
Abstracts
A. M. JACOBS: WHEN WILL WE EVER LEARN? MUSING ABOUT THE STATE OF THE ART IN MODELING BRAIN AND BEHAVIOR PROCESSES.
On the one hand, the recent explosive development of machine learning tools and parallel computing power has led to a plethora of cognitive and computational models of brain and behavior functions that can confuse many a researcher interested in theoretical psychology or neuroscience. Unfortunately, however, in most of this work in our field the hopeless strategy of 'one model for one task or dataset' prevails –badly ignoring our early proposals for functional overlap modeling and a general framework for building, testing, and evaluating computational models (Jacobs & Grainger, 1994, 1999). On the other hand, efforts to advance taxonomies or atlases for behavioral and brain domains only slowly take up speed and the blatant lack of agreement about the why, what and how of computational models and their testing continues to generate too many publications in too many journals without any real practical consequences (e.g., Cichy & Kaiser, 2019).
On the one hand, the recent explosive development of machine learning tools and parallel computing power has led to a plethora of cognitive and computational models of brain and behavior functions that can confuse many a researcher interested in theoretical psychology or neuroscience. Unfortunately, however, in most of this work in our field the hopeless strategy of 'one model for one task or dataset' prevails –badly ignoring our early proposals for functional overlap modeling and a general framework for building, testing, and evaluating computational models (Jacobs & Grainger, 1994, 1999). On the other hand, efforts to advance taxonomies or atlases for behavioral and brain domains only slowly take up speed and the blatant lack of agreement about the why, what and how of computational models and their testing continues to generate too many publications in too many journals without any real practical consequences (e.g., Cichy & Kaiser, 2019).
Similarly to Poldrack et al. (2019), here I argue that cognitive neuroscience should follow the lead taken in Data Science. This involves setting up internet repositories for both computational models and benchmark datasets (such as Hugging Face or GLUE). Unlike data science, however, our field can not focus too much on ‘benchmark boosting’, i.e. achieving ever increasing degrees of descriptive or predictive adequacy. Beyond that we also must publish and impose standards for evaluating horizontal and vertical generality, simplicity or falsifiability, and explanatory adequacy (Hofmann & Jacobs, 2014).
Similarly to Poldrack et al. (2019), here I argue that cognitive neuroscience should follow the lead taken in Data Science. This involves setting up internet repositories for both computational models and benchmark datasets (such as Hugging Face or GLUE). Unlike data science, however, our field can not focus too much on ‘benchmark boosting’, i.e. achieving ever increasing degrees of descriptive or predictive adequacy. Beyond that we also must publish and impose standards for evaluating horizontal and vertical generality, simplicity or falsifiability, and explanatory adequacy (Hofmann & Jacobs, 2014).
R. POLDRACK: WHY DO WE NEED A FORMAL ONTOLOGY OF COGNITION, AND WHAT SHOULD IT LOOK LIKE?
Cognitive ontologies have been primarily developed in a top-down manner on the basis of psychological theories that lean heavily upon folk psychological concepts. I will first assess the degree to which these ontologies are effective in predicting brain activity using cognitive encoding models, showing that they are surprisingly effective but also highlighting some important shortcomings. I will then ask whether we might develop better understanding of the organization of the mind through bottom-up, or “data driven” ontologies. I will present two examples of such data-driven ontologies, outlining their success but also detailing important limitations on this approach. I will end by discussing the need to envision ontologies based on computational rather than natural language descriptions.
Cognitive ontologies have been primarily developed in a top-down manner on the basis of psychological theories that lean heavily upon folk psychological concepts. I will first assess the degree to which these ontologies are effective in predicting brain activity using cognitive encoding models, showing that they are surprisingly effective but also highlighting some important shortcomings. I will then ask whether we might develop better understanding of the organization of the mind through bottom-up, or “data driven” ontologies. I will present two examples of such data-driven ontologies, outlining their success but also detailing important limitations on this approach. I will end by discussing the need to envision ontologies based on computational rather than natural language descriptions.
J. GOMEZ-LAVIN: MOSAIC ONTOLOGIES OF COGNITION: A TEST CASE FROM WORKING MEMORY.
The hope of an easy mapping of psychological function to neural structure has eroded, thanks in part to the advent of multivariate analyses demonstrating that no one region of the brain works in true isolation. Focusing on the case study of working memory—our famed ability to keep information in mind—I show that there is no coherent mapping from this psychological construct to a univocal neural structure. Instead, this failure prompts us to consider a strategy of productive pessimism when devising new, more explanatory mosaic ontologies of cognitive concepts.
The hope of an easy mapping of psychological function to neural structure has eroded, thanks in part to the advent of multivariate analyses demonstrating that no one region of the brain works in true isolation. Focusing on the case study of working memory—our famed ability to keep information in mind—I show that there is no coherent mapping from this psychological construct to a univocal neural structure. Instead, this failure prompts us to consider a strategy of productive pessimism when devising new, more explanatory mosaic ontologies of cognitive concepts.
Delay-period activity in the prefrontal cortex was long presumed to encode the contents of working memory. Challenging this dogma are lesion and electrophysiological studies disputing the necessity of prefrontal recruitment for working memory task performance. Additionally, the discovery of stimulus-bounded delay-period activity throughout the cortex and so-called “activity silent” working memory paradigms demonstrate that the brain possesses a plurality of mechanisms charged with maintaining and manipulating information.
Delay-period activity in the prefrontal cortex was long presumed to encode the contents of working memory. Challenging this dogma are lesion and electrophysiological studies disputing the necessity of prefrontal recruitment for working memory task performance. Additionally, the discovery of stimulus-bounded delay-period activity throughout the cortex and so-called “activity silent” working memory paradigms demonstrate that the brain possesses a plurality of mechanisms charged with maintaining and manipulating information.
How do we advance from this pessimistic position? Should we relegate working memory to the class of flawed functionalist concepts, akin to a modern phlogiston? Despite the evidence against working memory’s status as an explanatory and coherent natural kind, it’s clear that we can keep information in mind, and that this process is central to many cognitions. I argue that working memory and other cognitive concepts can play productive roles in new ontologies if we abandon a strong commitment to their explanatory role while also privileging their organizational utility. There is not one revelatory way that we keep information in mind, but several, many of which are already described (and conflated) within the literature under the label ‘working memory.’ Privileging the mosaic description underlying cognitive concepts may yield more explanatory ontologies of cognition.
How do we advance from this pessimistic position? Should we relegate working memory to the class of flawed functionalist concepts, akin to a modern phlogiston? Despite the evidence against working memory’s status as an explanatory and coherent natural kind, it’s clear that we can keep information in mind, and that this process is central to many cognitions. I argue that working memory and other cognitive concepts can play productive roles in new ontologies if we abandon a strong commitment to their explanatory role while also privileging their organizational utility. There is not one revelatory way that we keep information in mind, but several, many of which are already described (and conflated) within the literature under the label ‘working memory.’ Privileging the mosaic description underlying cognitive concepts may yield more explanatory ontologies of cognition.
K. ROBBINS: HED - MOVING BEYOND KEYWORDS TO FINE-GRAINED ANALYSIS-READY ANNOTATION IN NEUROIMAGING.
A substantial gap exists between the level of description required by current digital data archiving standards and the level of annotation required for successful analysis of event-related and other time-series data across studies, environments, and laboratories. The new third-generation formulation of the Hierarchical Event Descriptors (HED-3G) framework and tools (hedtags.org) provides the infrastructure needed to document events and their interrelationships, resulting in machine-actionable annotation for automated within- and across-study comparisons and analysis. Powerful annotation strategies combine robust event description with details of experiment design, experimental tasks, participant actions, and dataset metadata to produce fine-grained annotation in human-readable as well as machine-actionable form, making event annotation relevant to the full range of neuroimaging and other time series data. In this talk, I will demonstrate how researchers can use HED-3G to define higher-level “concepts” that map their local coding and notations into a machine-actionable framework using HED vocabularies or community-developed library vocabularies. HED has been incorporated into the BIDS (Brain Imaging Data Structure) specification and validation framework. For BIDS datasets, adding a single JSON events sidecar file or improving an existing one may be all that is needed to turn an unusable dataset into a richly informative one.
A substantial gap exists between the level of description required by current digital data archiving standards and the level of annotation required for successful analysis of event-related and other time-series data across studies, environments, and laboratories. The new third-generation formulation of the Hierarchical Event Descriptors (HED-3G) framework and tools (hedtags.org) provides the infrastructure needed to document events and their interrelationships, resulting in machine-actionable annotation for automated within- and across-study comparisons and analysis. Powerful annotation strategies combine robust event description with details of experiment design, experimental tasks, participant actions, and dataset metadata to produce fine-grained annotation in human-readable as well as machine-actionable form, making event annotation relevant to the full range of neuroimaging and other time series data. In this talk, I will demonstrate how researchers can use HED-3G to define higher-level “concepts” that map their local coding and notations into a machine-actionable framework using HED vocabularies or community-developed library vocabularies. HED has been incorporated into the BIDS (Brain Imaging Data Structure) specification and validation framework. For BIDS datasets, adding a single JSON events sidecar file or improving an existing one may be all that is needed to turn an unusable dataset into a richly informative one.
J. TURNER: BRIDGING THE GAP FROM PUBLICATION TO PRACTICE IN NEUROIMAGING DATA SHARING.
As a field, neuroimaging researchers have been engaged with data sharing techniques for over 20 years. The Cognitive Paradigm Ontology (CogPO) focused on representing the experimental techniques used in neuroimaging that support conclusions regarding brain/behavior relationships, with the goal of facilitating aggregated analyses of the literature. While incomplete, it showed the usefulness and promise of structured terminologies in understanding the breadth of cognitive neuroscience publications. Similarly, the SchizConnect project developed federated access to several public data repositories, with the goal of facilitating aggregated analyses of the datasets. To link the user to the data required standard terminologies that could reflect the variety of study design and variables, necessitating basic conceptual structures. The terminologies developed in the one case based on the literature and on the other from the need to represent dataset characteristics reflect differing needs and constraints; but both required significant investment of human input to take shape, highlighting the need for powerful and flexible automated methods for growing and modifying any conceptual representations. The final project I will discuss, NeuroBridge, builds on both these efforts and others to facilitate finding relevant datasets from both data repositories and the literature.
As a field, neuroimaging researchers have been engaged with data sharing techniques for over 20 years. The Cognitive Paradigm Ontology (CogPO) focused on representing the experimental techniques used in neuroimaging that support conclusions regarding brain/behavior relationships, with the goal of facilitating aggregated analyses of the literature. While incomplete, it showed the usefulness and promise of structured terminologies in understanding the breadth of cognitive neuroscience publications. Similarly, the SchizConnect project developed federated access to several public data repositories, with the goal of facilitating aggregated analyses of the datasets. To link the user to the data required standard terminologies that could reflect the variety of study design and variables, necessitating basic conceptual structures. The terminologies developed in the one case based on the literature and on the other from the need to represent dataset characteristics reflect differing needs and constraints; but both required significant investment of human input to take shape, highlighting the need for powerful and flexible automated methods for growing and modifying any conceptual representations. The final project I will discuss, NeuroBridge, builds on both these efforts and others to facilitate finding relevant datasets from both data repositories and the literature.
G. GUIZZARDI: OBJECT META-TYPES, CROSS-WORLD IDENTITY AND CONCEPTUAL SPACES.
G. GUIZZARDI: OBJECT META-TYPES, CROSS-WORLD IDENTITY AND CONCEPTUAL SPACES.
Types are fundamental for domain modeling, ontology engineering, and knowledge representation in computer science. Frequently, monadic types used in these models have as instances objects (endurants, continuants), i.e., entities persisting in time that experience qualitative changes while keeping their numerical identity. In this talk, I present a descriptive formal ontology of endurant meta-types (e.g., kinds, phases, roles, mixins). Moreover, I also show here how this proposal can complement the theory of conceptual spaces by offering an account for kind-supplied principles of cross-world identity. The account addresses an important criticism posed to conceptual spaces in the literature and is in line with a number of empirical results in the literature of cognitive psychology. Finally, I will briefly discuss some logical systems designed to formally characterize the distinctions and constraints proposed by this theory.
Types are fundamental for domain modeling, ontology engineering, and knowledge representation in computer science. Frequently, monadic types used in these models have as instances objects (endurants, continuants), i.e., entities persisting in time that experience qualitative changes while keeping their numerical identity. In this talk, I present a descriptive formal ontology of endurant meta-types (e.g., kinds, phases, roles, mixins). Moreover, I also show here how this proposal can complement the theory of conceptual spaces by offering an account for kind-supplied principles of cross-world identity. The account addresses an important criticism posed to conceptual spaces in the literature and is in line with a number of empirical results in the literature of cognitive psychology. Finally, I will briefly discuss some logical systems designed to formally characterize the distinctions and constraints proposed by this theory.
B. SMITH: FROM BFO TO MFO: A MODULAR APPROACH TO COGNITIVE ONTOLOGIES.
As experts from diverse fields of research come together in the development of cognitive ontologies to aid knowledge accumulation and integration in cognitive neuroscience, it is important that a way is found to ensure the interoperability of the ontologies which result.
As experts from diverse fields of research come together in the development of cognitive ontologies to aid knowledge accumulation and integration in cognitive neuroscience, it is important that a way is found to ensure the interoperability of the ontologies which result.
Fatefully, the biomedical community found a way of avoiding the ontology silo problem almost from the very start, through the creation of the Open Biological and Biomedical Ontologies (OBO) Foundry (https://www.nature.com/articles/nbt1346), a consortium of ontology developers who committed to using a single hub and spokes approach in order to enable the creation of interoperable ontology modules. The central hub in this architecture is Basic Formal Ontology (BFO), which encapsulates in its structure a number of principles designed to enable successful development of interoperable ontology modules (https://mitpress.mit.edu/books/building-ontologies-basic-formal-ontology). These principles have been tested and refined over almost two decades in the work of the OBO Foundry, and they are increasingly being adopted in other spheres. This has led in turn to a process on the part of the International Standards Organization and the International Electrotechnical Commission to create standard ISO/IEC 21838, part 1 of which lays down the requirements which must be satisfied by a top-level ontology, while part 2 certifies that Basic Formal Ontology satisfies these requirements.
Fatefully, the biomedical community found a way of avoiding the ontology silo problem almost from the very start, through the creation of the Open Biological and Biomedical Ontologies (OBO) Foundry (https://www.nature.com/articles/nbt1346), a consortium of ontology developers who committed to using a single hub and spokes approach in order to enable the creation of interoperable ontology modules. The central hub in this architecture is Basic Formal Ontology (BFO), which encapsulates in its structure a number of principles designed to enable successful development of interoperable ontology modules (https://mitpress.mit.edu/books/building-ontologies-basic-formal-ontology). These principles have been tested and refined over almost two decades in the work of the OBO Foundry, and they are increasingly being adopted in other spheres. This has led in turn to a process on the part of the International Standards Organization and the International Electrotechnical Commission to create standard ISO/IEC 21838, part 1 of which lays down the requirements which must be satisfied by a top-level ontology, while part 2 certifies that Basic Formal Ontology satisfies these requirements.
I will describe how the Mental Functioning Ontology (MFO) was created in accordance with the OBO Foundry principles, and how MFO has been extended to create the Emotion Ontology (MFO-EM) and the Mental Disease Ontology (MFO-MD). I will also comment on current work applying MFO to the topic of mental capabilities.
I will describe how the Mental Functioning Ontology (MFO) was created in accordance with the OBO Foundry principles, and how MFO has been extended to create the Emotion Ontology (MFO-EM) and the Mental Disease Ontology (MFO-MD). I will also comment on current work applying MFO to the topic of mental capabilities.
A. RAVENSCHLAG: CAN A MULTI-MODULAR ONTOLOGY AID PROGRESS IN COGNITIVE NEUROSCIENCE?
Establishing effective data management and sharing practices as hallmarks for progress in open science has received increasing attention within the recent years. Having said that, Cognitive Neuroscience, like many other disciplines, is still facing the challenge of integrating its conceptual knowledge with the vastly growing amounts of (neuroimaging) data. To fully exploit our data’s value and potential, we need to find a way to structure our scientific knowledge so that it can serve as a basis for rich data annotation and thus data (re-)usability. Seminal efforts towards this goal have already been made with collaborative knowledge building projects (e.g. Cognitive Atlas) and objective experiment description systems (e.g. CogPO). Given the latent and only indirectly observable nature of cognitive constructs, we aim to contribute to these efforts by introducing the idea of a theory-based, multi-modular cognitive ontology. Cognitive concepts as such are rooted in the context of their underlying theories, potentially leading to multiple definitions for terminologically identical concepts. Therefore, we argue that incorporating cognitive theories and models as guiding principles for knowledge representation could allow for the coexistence of diverging definitions and thus might bear the potential for comprehensive ontology-based data indexing, access, and reasoning. I will give an overview of the modular structure of our envisioned knowledge representation system, including functional, theoretical, experimental, and analytical aspects of neurocognitive research - meant to provide maximal flexibility regarding conceptualisation, maintenance, and applicability from both epistemological and user perspectives.
Establishing effective data management and sharing practices as hallmarks for progress in open science has received increasing attention within the recent years. Having said that, Cognitive Neuroscience, like many other disciplines, is still facing the challenge of integrating its conceptual knowledge with the vastly growing amounts of (neuroimaging) data. To fully exploit our data’s value and potential, we need to find a way to structure our scientific knowledge so that it can serve as a basis for rich data annotation and thus data (re-)usability. Seminal efforts towards this goal have already been made with collaborative knowledge building projects (e.g. Cognitive Atlas) and objective experiment description systems (e.g. CogPO). Given the latent and only indirectly observable nature of cognitive constructs, we aim to contribute to these efforts by introducing the idea of a theory-based, multi-modular cognitive ontology. Cognitive concepts as such are rooted in the context of their underlying theories, potentially leading to multiple definitions for terminologically identical concepts. Therefore, we argue that incorporating cognitive theories and models as guiding principles for knowledge representation could allow for the coexistence of diverging definitions and thus might bear the potential for comprehensive ontology-based data indexing, access, and reasoning. I will give an overview of the modular structure of our envisioned knowledge representation system, including functional, theoretical, experimental, and analytical aspects of neurocognitive research - meant to provide maximal flexibility regarding conceptualisation, maintenance, and applicability from both epistemological and user perspectives.
M. DENISSEN: A GATEWAY TO DATA: EXPERIMENTS AS A POTENTIAL HUB BETWEEN CONCEPTUAL KNOWLEDGE AND NEUROIMAGING DATA.
M. DENISSEN: A GATEWAY TO DATA: EXPERIMENTS AS A POTENTIAL HUB BETWEEN CONCEPTUAL KNOWLEDGE AND NEUROIMAGING DATA.
While there is growing consensus on the need of data and metadata harmonization to advance open data sharing and (re-)use, the question remains of how to effectively integrate data content and context. In Cognitive Neuroscience, the experiment determines the cognitive state of a participant, which is what ultimately gives meaning to our data. Standardisation of experiment descriptions are already under way and these efforts are key to integrating neurocognitive correlates with semantic knowledge representations. Critically, while the experiment clearly paves the way on what research questions can be addressed by a given dataset, answering these questions depends on the analysis, which in turn informs cognitive theory. In consequence, meta-data on analysis characteristics needs to be distinguishable from meta-data describing experimental characteristics. Furthermore, experiment descriptions need to provide a reasonably fine level of granularity. We therefore suggest that the connection between data and conceptual knowledge can be best captured by adding a comprehensive analysis description on top of currently available tools for standardised data structuring (e.g., BIDS) and event description (e.g., HED). This way, while the experimental module of our envisioned ontology would maximise data (re-)usability, the analysis module would not only enable us to foster our understanding of cognitive concepts in a data-driven fashion, but also to encourage secondary analysis, while allowing for integrating diverging theoretical perspectives into the knowledge system.
While there is growing consensus on the need of data and metadata harmonization to advance open data sharing and (re-)use, the question remains of how to effectively integrate data content and context. In Cognitive Neuroscience, the experiment determines the cognitive state of a participant, which is what ultimately gives meaning to our data. Standardisation of experiment descriptions are already under way and these efforts are key to integrating neurocognitive correlates with semantic knowledge representations. Critically, while the experiment clearly paves the way on what research questions can be addressed by a given dataset, answering these questions depends on the analysis, which in turn informs cognitive theory. In consequence, meta-data on analysis characteristics needs to be distinguishable from meta-data describing experimental characteristics. Furthermore, experiment descriptions need to provide a reasonably fine level of granularity. We therefore suggest that the connection between data and conceptual knowledge can be best captured by adding a comprehensive analysis description on top of currently available tools for standardised data structuring (e.g., BIDS) and event description (e.g., HED). This way, while the experimental module of our envisioned ontology would maximise data (re-)usability, the analysis module would not only enable us to foster our understanding of cognitive concepts in a data-driven fashion, but also to encourage secondary analysis, while allowing for integrating diverging theoretical perspectives into the knowledge system.
Slides and Recordings
Slides and Recordings
Slides and recordings of the ISCO talks can now be downloaded via the following links:
A. M. JACOBS: WHEN WILL WE EVER LEARN? MUSING ABOUT THE STATE OF THE ART IN MODELING BRAIN AND BEHAVIOR PROCESSES.
(VIDEO)
A. M. JACOBS: WHEN WILL WE EVER LEARN? MUSING ABOUT THE STATE OF THE ART IN MODELING BRAIN AND BEHAVIOR PROCESSES.
(VIDEO)
R. POLDRACK: WHY DO WE NEED A FORMAL ONTOLOGY OF COGNITION, AND WHAT SHOULD IT LOOK LIKE? (VIDEO) (SLIDES)
SHORT QUESTIONS (VIDEO)
J. GOMEZ-LAVIN: MOSAIC ONTOLOGIES OF COGNITION: A TEST CASE FROM WORKING MEMORY. (VIDEO) (SLIDES)
SHORT QUESTIONS & INTERIM DISCUSSION (VIDEO)
K. ROBBINS: HED - MOVING BEYOND KEYWORDS TO FINE-GRAINED ANALYSIS-READY ANNOTATION IN NEUROIMAGING. (VIDEO) (SLIDES)
SHORT QUESTIONS (VIDEO)
J. TURNER: BRIDGING THE GAP FROM PUBLICATION TO PRACTICE IN NEUROIMAGING DATA SHARING. (VIDEO) (SLIDES)
SHORT QUESTIONS & INTERIM DISCUSSION (VIDEO)
G. GUIZZARDI: OBJECT META-TYPES, CROSS-WORLD IDENTITY AND CONCEPTUAL SPACES. (VIDEO) (SLIDES)
SHORT QUESTIONS (VIDEO)
B. SMITH: FROM BFO TO MFO: A MODULAR APPROACH TO COGNITIVE ONTOLOGIES. (VIDEO) (SLIDES)
SHORT QUESTIONS & INTERIM DISCUSSION (VIDEO)
A. RAVENSCHLAG: CAN A MULTI-MODULAR ONTOLOGY AID PROGRESS IN COGNITIVE NEUROSCIENCE? (VIDEO) (SLIDES)
M. DENISSEN: A GATEWAY TO DATA: EXPERIMENTS AS A POTENTIAL HUB BETWEEN CONCEPTUAL KNOWLEDGE AND NEUROIMAGING DATA. (VIDEO) (SLIDES)
PANEL DISCUSSION (VIDEO)
Contact
Contact