GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter).
Objectives To assess a range of tools and methods to support Trusted Research Environments (TREs) to assess output from AI methods for potentially identifiable information, investigate the legal and ethical implications and controls, and produce a set of guidelines and recommendations to support all...
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Swansea University
2022-08-01
|
| Series: | International Journal of Population Data Science |
| Subjects: | |
| Online Access: | https://ijpds.org/article/view/2005 |
| _version_ | 1797430640311795712 |
|---|---|
| author | Emily Jefferson Christian Cole Alba Crespi Boixader Simon Rogers Maeve Malone Felix Ritchie Jim Smith Francesco Tava Angela Daly Jillian Beggs Antony Chuter |
| author_facet | Emily Jefferson Christian Cole Alba Crespi Boixader Simon Rogers Maeve Malone Felix Ritchie Jim Smith Francesco Tava Angela Daly Jillian Beggs Antony Chuter |
| author_sort | Emily Jefferson |
| collection | DOAJ |
| description | Objectives
To assess a range of tools and methods to support Trusted Research Environments (TREs) to assess output from AI methods for potentially identifiable information, investigate the legal and ethical implications and controls, and produce a set of guidelines and recommendations to support all TREs with export controls of AI algorithms.
Approach
TREs provide secure facilities to analyse confidential personal data, with staff checking outputs for disclosure risk before publication. Artificial intelligence (AI) has high potential to improve the linking and analysis of population data, and TREs are well suited to supporting AI modelling. However, TRE governance focuses on classical statistical data analysis. The size and complexity of AI models presents significant challenges for the disclosure-checking process. Models may be susceptible to external hacking: complicated methods to reverse engineer the learning process to find out about the data used for training, with more potential to lead to re-identification than conventional statistical methods.
Results
GRAIMatter is:
• Quantitatively assessing the risk of disclosure from different AI models exploring different models, hyper-parameter settings and training algorithms over common data types
• Evaluating a range of tools to determine effectiveness for disclosure control
• Assessing the legal and ethical implications of TREs supporting AI development and identifying aspects of existing legal and regulatory frameworks requiring reform.
• Running 4 PPIE workshops to understand their priorities and beliefs around safeguarding and securing data
• Developing a set of recommendations including
• suggested open-source toolsets for TREs to use to measure and reduce disclosure risk
• descriptions of the technical and legal controls and policies TREs should implement across the 5 Safes to support AI algorithm disclosure control
• training implications for both TRE staff and how they validate researchers
Conclusion
GRAIMatter is developing a set of usable recommendations for TREs to use to guard against the additional risks when disclosing trained AI models from TREs.
|
| first_indexed | 2024-03-09T09:31:28Z |
| format | Article |
| id | doaj.art-edf4041fad1b431f8adc5d9fc067019f |
| institution | Directory Open Access Journal |
| issn | 2399-4908 |
| language | English |
| last_indexed | 2024-03-09T09:31:28Z |
| publishDate | 2022-08-01 |
| publisher | Swansea University |
| record_format | Article |
| series | International Journal of Population Data Science |
| spelling | doaj.art-edf4041fad1b431f8adc5d9fc067019f2023-12-02T03:57:44ZengSwansea UniversityInternational Journal of Population Data Science2399-49082022-08-017310.23889/ijpds.v7i3.2005GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter).Emily Jefferson0Christian Cole1Alba Crespi Boixader2Simon Rogers3Maeve Malone 4Felix Ritchie5Jim Smith6Francesco Tava7Angela Daly8Jillian Beggs9Antony Chuter10University of DundeeUniversity of DundeeUniversity of DundeeNHS ScotlandUniversity of DundeeUniversity of West of EnglandUniversity of West of EnglandUniversity of West of EnglandUniversity of DundeePPIE Co-IPPIE Co-IObjectives To assess a range of tools and methods to support Trusted Research Environments (TREs) to assess output from AI methods for potentially identifiable information, investigate the legal and ethical implications and controls, and produce a set of guidelines and recommendations to support all TREs with export controls of AI algorithms. Approach TREs provide secure facilities to analyse confidential personal data, with staff checking outputs for disclosure risk before publication. Artificial intelligence (AI) has high potential to improve the linking and analysis of population data, and TREs are well suited to supporting AI modelling. However, TRE governance focuses on classical statistical data analysis. The size and complexity of AI models presents significant challenges for the disclosure-checking process. Models may be susceptible to external hacking: complicated methods to reverse engineer the learning process to find out about the data used for training, with more potential to lead to re-identification than conventional statistical methods. Results GRAIMatter is: • Quantitatively assessing the risk of disclosure from different AI models exploring different models, hyper-parameter settings and training algorithms over common data types • Evaluating a range of tools to determine effectiveness for disclosure control • Assessing the legal and ethical implications of TREs supporting AI development and identifying aspects of existing legal and regulatory frameworks requiring reform. • Running 4 PPIE workshops to understand their priorities and beliefs around safeguarding and securing data • Developing a set of recommendations including • suggested open-source toolsets for TREs to use to measure and reduce disclosure risk • descriptions of the technical and legal controls and policies TREs should implement across the 5 Safes to support AI algorithm disclosure control • training implications for both TRE staff and how they validate researchers Conclusion GRAIMatter is developing a set of usable recommendations for TREs to use to guard against the additional risks when disclosing trained AI models from TREs. https://ijpds.org/article/view/2005AIArtificial IntelligenceMachine LearningTrusted Research EnvironmentsSafe Haven EnvironmentsDisclosure Control |
| spellingShingle | Emily Jefferson Christian Cole Alba Crespi Boixader Simon Rogers Maeve Malone Felix Ritchie Jim Smith Francesco Tava Angela Daly Jillian Beggs Antony Chuter GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). International Journal of Population Data Science AI Artificial Intelligence Machine Learning Trusted Research Environments Safe Haven Environments Disclosure Control |
| title | GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). |
| title_full | GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). |
| title_fullStr | GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). |
| title_full_unstemmed | GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). |
| title_short | GRAIMatter: Guidelines and Resources for AI Model Access from TrusTEd Research environments (GRAIMatter). |
| title_sort | graimatter guidelines and resources for ai model access from trusted research environments graimatter |
| topic | AI Artificial Intelligence Machine Learning Trusted Research Environments Safe Haven Environments Disclosure Control |
| url | https://ijpds.org/article/view/2005 |
| work_keys_str_mv | AT emilyjefferson graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT christiancole graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT albacrespiboixader graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT simonrogers graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT maevemalone graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT felixritchie graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT jimsmith graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT francescotava graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT angeladaly graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT jillianbeggs graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter AT antonychuter graimatterguidelinesandresourcesforaimodelaccessfromtrustedresearchenvironmentsgraimatter |