Technology Glossary
Published: 25 September 2023
6 minute read
Within the technology space, there are a great number of sometimes competing definitions that attempt to describe the many different types of technology currently in use or development. When engaging with stakeholders on the use of technology, it is therefore important that the FRC is consistent and clear across projects and is able to articulate the meaning behind key terms.
There is great benefit in establishing a common language both within the FRC and with our key external stakeholders. A common language ensures that all participants in conversations understand each other, misunderstandings are minimised, and a sense of consistency can be created during engagement and in material published by the FRC.
In defining the terms below, we are able to clarify the concepts underlying some of these key terminologies and ensure that the FRC has a functional understanding of these technologies. These definitions encompass current thinking and are designed to be understandable by non-expert users. They do not provide in-depth knowledge or attempt to summarise the entirety of many complex fields of research and development. Where the FRC is in need of such detailed understanding, we should consult with experts in the respective fields and seek to address any knowledge gaps through additional training.
These definitions are likely to evolve over time, and do not represent the only possible way of describing these terms, but they provide a sensible starting point for conversations with stakeholders. We will review these definitions on a timely basis as the use of technology evolves to ensure they are still reflective of current thinking, and accurately reflect the technologies they are being used to describe.
| Term | Resource | Definition | Category |
|---|---|---|---|
| Artificial Intelligence (AI) | The Alan Turing Institute | In computer science, the term artificial intelligence (AI) refers to any human-like intelligence exhibited by a computer, robot, or other machine. In popular usage, artificial intelligence refers to the ability of a computer or machine to mimic the capabilities of the human mind—learning from examples and experience, recognizing objects, understanding, and responding to language, making decisions, solving problems—and combining these and other capabilities to perform functions a human might perform. | Artificial Intelligence & Related Terms |
| Black-Box Tools | Nature | Black-box AI tools have observable inputs and outputs but lack clarity on the inner workings of the tool, it may identify an item as riskier but there isn’t necessarily an easily discernible “how” or “why” in the process. | Artificial Intelligence & Related Terms |
| Explainability | The Alan Turing Institute | ‘Explainability’ is the second stage in the process of describing the workings of an AI tool, after interpretability. Black-box AI tools for example, which are not particularly interpretable from the outset, require detailed post-hoc analysis in order to explain how a tool has produced a given result.Though complex, the inner workings of such an AI tool can be explained through post-hoc analysis – where date sets are created, fed into the model, and adjusted with the subsequent outputs recorded and analysed to develop an understanding of the process occurring within the model. Though this is possible, it is a far more complex process than creating models which are inherently interpretable, and thus easier to explain, in the first instance. Visualisation is a key tool in explaining AI tools and their outputs. | Artificial Intelligence & Related Terms |
| Interpretability | The Alan Turing Institute | ‘Interpretability’ describes how intrinsically understandable an AI tool is before the application of any technique to explain the tools workings. Simple AI tools which make use of more simple statistical models and techniques generally have a higher degree of interpretability than complex models using more advanced techniques | Artificial Intelligence & Related Terms |
| Intrinsically Interpretable AI | Nature | Intrinsically Interpretable tools have observable inputs and outputs as with a black box AI tool, but also understandable behaviours and relationships in arriving at those outputs. Some of these types of tools come an associated cost of lower accuracy, though such tools may be as accurate as black box AI tools if designed using certain techniques.In the context of AI, intrinsically interpretable AI refers to methods used in AI application such that the results of the process can be understood. It contrasts with the "black box" approach to machine learning where even an algorithms designers will encounter difficulty in understanding the algorithms processes. | Artificial Intelligence & Related Terms |
| Justifiability | Internally Developed Definition | ‘Justifiability’ is the ability to justify an algorithms use in any specific context including, for example, why it was acceptable to apply any given tool to a dataset and how the users ensured the data input was of sufficient quality. Here, we are describing the process of justifying a tools use in a particular scenario, such as to aid in risk assessment during an audit process, as oppose to justifying why any given algorithm was used over another. | Artificial Intelligence & Related Terms |
| Machine Learning | The Alan Turing Institute | Machine learning is a subset of AI application that learns by itself. It reprograms itself as it digests more data, allowing it to perform the specific task it's designed to perform with increasingly greater accuracy. | Artificial Intelligence & Related Terms |
| Natural Language Processing | The Alan Turing Institute | Natural language processing (NLP) refers to the branch of AI concerned with creating computer systems that are able to understand spoken word and text inputs in a way similar to humans.NLP combines a rule-based model of human language with statistical, machine learning, and deep learning models. It is not, for example, a chatbot simply searching for terms in a database once a user has input them, it involves the tool deriving the intent and sentiment of the text or voice data in order to understand what has been input and what the user is looking to achieve with the process. | Artificial Intelligence & Related Terms |
| Robotic Process Automation | IBM | Robotic process automation (RPA) is a type of automation which makes use of software robots, commonly referred to as bots, though RPA may also make use of AI.Generally, RPA systems automate by observing a user perform a task, such as transferring information between systems, and then performs the automation by repeating the task. A process of refinement is required in order to ensure that the RPA is able to perform the task to a high degree of accuracy. | Artificial Intelligence & Related Terms |
| Automated Tools and Techniques (ATT) | ISA (UK) 315 (Revised July 2020), paragraph A21 onwards | Technology used to perform risk assessment procedures and / or obtain audit evidence. A subset of technological resources. | Auditing Terminology |
| Audit data analytics (ADA) | As used in AQR’s 2020 review, taken from the IAASB Data Analytics Working Group’s Request for Input dated September 2016 | A subset of Automated Tools and Techniques.“The science and art of discovering and analysing patterns, deviations and inconsistencies and identifying anomalies, and extracting other useful information in data underlying or related to the subject of an audit through analysis, modelling and visualisation for the purpose of planning or performing the audit.”An ADA or ADAs are data analytic techniques that can be used to perform risk assessment, tests of controls, substantive procedures (that is tests of details or substantive analytical procedures) or concluding audit procedures.For clarity, we do not use the term ADA to refer to automated tools and techniques that involve the use of artificial intelligence (AI) or machine learning (ML). | Auditing Terminology |
| Technological Resources | ISA 220 (Revised 2020), paragraphs A63 – A67 | Umbrella term for technology that assists the auditor performing risk assessment procedures, obtaining audit evidence and / or managing the audit process. | Auditing Terminology |
| Blockchain & Distributed Ledger Technology | Internally Developed Definition | A blockchain is a type of database used to record information.A traditional database organises data into tables, and a single copy of that database is generally held on one computer system or by a single individual. Blockchain however organises its data into chunks, referred to as blocks, that are chained together to create a timeline of data that is duplicated and distributed across the entire network of computer systems on the blockchain. Participants in the network are referred to as nodes.Each block in the chain contains a number of transactions, and every time a new transaction occurs on the blockchain, a record of that transaction is added to every participant’s copy of the chain, but existing data is not modified. Blocks are added to the chain by solving complex cryptographic problems. This type of blockchain is known as a ‘proof of work’ chain.This decentralised approach to a database means that manipulation of data on the chain would require more than 50% of nodes to be altered in order for changes to be accepted, a huge undertaking, and one which requires significant computing power.A ‘proof of stake’ chain uses most of the same principles, but in order to add a new block to the chain, validators do not solve complex cryptographic equations in the same way. Instead, they participate in the network through holding a stake in the network itself, for example by holding tokens or coins, with returns for staking their coins awarded based on the level of investment. Attempts to hack or manipulate the network result in the loss of staked coins.Where the block chain is set up to work in this way it is a decentralised database managed by multiple participants and is known as Distributed Ledger Technology (DLT). Ledger is used here in a broader sense than it is generally used in accounting. Private, centralised databases can be created which operate in a similar way, and retain the benefits associated with blockchain technology, but all of the computers which hold a copy of the ledger are managed by a single participant or entity. | Blockchain & Cryptocurrency |
| Cryptocurrency | Internally Developed Definition | Cryptocurrencies are digital assets (often referred to as coins) which can be used as an exchange medium. They may be used to settle transactions, similar to how traditional currency is used, but generally without the need for a central issuing authority or bank.Instead, blockchain technology is used to create a decentralised control network, with cryptography used to verify the transfer of coin ownership. Networks may also be created that rely on pre-mined or “minted” cryptocurrency, which operate in a centralised way more similar to a traditional currency created by a central issuing authority. | Blockchain & Cryptocurrency |
| Common Data Model | Microsoft | Common data models are models which organise elements of data, providing a standard system and set of rules around how those elements of data relate to each other.They are generally designed to create a shared language between multiple systems and applications, allowing data to easily be shared between them, facilitating business and analytical applications. | Other Definitions |