Request ID: FOI-0487-2324 Date published: 14 June 2023
You asked
Please publish your policies and standards for the evaluation of AI in decision support services or systems.
For clarity by "AI" I mean the term in the same way that it is defined in the proposed EU AI Act: ‘Any software that is developed with one or more of the techniques and approaches listed in Annex I and can, for a given set of human-defined objectives, generate outputs, such as content, predictions, recommendations, or decisions influencing the environments they interact with.’ where Annex 1 envisages three specific techniques and approaches for AI: 'machine learning approaches', including supervised, unsupervised and reinforcement learning, using a wide variety of methods including deep learning; 'logic and knowledge-based approaches', including knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, (symbolic) reasoning and expert systems; and, 'statistical approaches', Bayesian estimation, search and optimisation methods.
We answered
TfL Ref: 0487-2324
Thank you for your request received by Transport for London (TfL) on 18 May 2023 asking for information about Artificial Intelligence (AI).
Your request has been considered in accordance with the requirements of the Freedom of Information Act and our information access policy. I can confirm that we hold the information you require. You asked: Please publish your policies and standards for the evaluation of AI in decision support services or systems.
For clarity by "AI" I mean the term in the same way that it is defined in the proposed EU AI Act: ‘Any software that is developed with one or more of the techniques and approaches listed in Annex I and can, for a given set of human-defined objectives, generate outputs, such as content, predictions, recommendations, or decisions influencing the environments they interact with.’ where Annex 1 envisages three specific techniques and approaches for AI: 'machine learning approaches', including supervised, unsupervised and reinforcement learning, using a wide variety of methods including deep learning; 'logic and knowledge-based approaches', including knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, (symbolic) reasoning and expert systems; and, 'statistical approaches', Bayesian estimation, search and optimisation methods.
We provide advice and instructions, as part of TfL’s Management System, to all staff on Profiling and Automated Decision Making in relation to personal data, which covers some aspects of AI: The relevant pages from the Management System (an online tool) are reproduced in the attached.
There are questions relating to AI in our Data Protection Impact Assessment (DPIA) template which is also attached.
If this is not the information you are looking for, or if you are unable to access it for any reason, please do not hesitate to contact me.
Please see the attached information sheet for details of your right to appeal as well as information on copyright and what to do if you would like to re-use any of the information we have disclosed.
Yours sincerely
Sara Thomas FOI Case Management Team General Counsel Transport for London
