If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Ethical artificial intelligence (AI) frameworks can be the catalyst in improving the safety and wellbeing of people when developing AI systems. In 2020 Rolls-Royce released its ethical and trustworthiness toolkit, The Aletheia FrameworkTM, which helps guide organisations as they consider the ethics around the use of AI. It covers three facets: social impact, accuracy and trust, and governance - which apply across all uses of AI. By adopting AI ethics and trust frameworks, oncologists can ensure the ratio between the benefit and harms of AI can be maximised. With AI transforming every sector, collaboration across industries to share ideas and learn from each other - even unlikely partnerships between engineering and oncology - could help optimise that transition.
The ethical deployment and governance of artificial intelligence are crucial to enable the widescale deployment of artificial intelligence systems needed to improve the wellbeing and safety of people, but in many domains the pace at which artificial intelligence is developing outstrips the pace at which regulations are being developed. Artificial intelligence systems should improve the wellbeing and safety of people. However, there are many examples where artificial intelligence-based systems have been shown to reproduce existing biases (e.g. racial biases in facial recognition and court sentencing decisions). There are several frameworks that provide theoretical guidance to these, and other, challenges, but methods to practically apply them are rare.
To help contribute to solving these challenges and given its responsibility as a trusted provider in the aerospace industry, Rolls-Royce released a pioneering ethics toolkit and trustworthiness process called The Aletheia Framework™ [
]. It focuses on moving beyond theory and into the practical application of responsible and trustworthy artificial intelligence: going from the ‘what?’ to the ‘how?’. The framework also allows artificial intelligence developers to tackle prevailing issues in automated systems, including preventing the reinforcement of social biases, sustaining the jobs and skills of people, covering accountability to ensuring trust in the outputs of an algorithm and more.
Rolls-Royce's power technologists and oncology professionals are experts in their respective domains. The public is frequently the end user and both domains need the public to trust their decision making, quality, products and services in order to keep them safe. This article will discuss how a toolkit from the power sector is relevant to cancer care.
Rolls-Royce parted ways from the motorcar industry in the 1990s and is now a global power group where digital technology is the key to business in safety or business-critical contexts and in extreme environments. As well as operating within demanding regulatory frameworks, the company has rigorous internal processes around product safety. With artificial intelligence being used daily in Rolls-Royce throughout the lifecycle of products, that same level of product safety rigour being used in designing and operating such things as its jet engines is used in its treatment and use of data, artificial intelligence and analytics. In contrast to the aerospace industry, commercial artificial intelligence products have only recently been introduced in radiotherapy clinics, and efforts have focused on demonstrating performance, either in academic or clinical settings, and product approval. To date, commercial artificial intelligence tools for radiotherapy are released as static products, whose performance can be monitored by oncology professionals. An agile lifecycle management process (e.g. where artificial intelligence-based segmentation models are consistently updated with new patient data) may be of interest, but is unlikely to be available in the near future.
An example of Rolls-Royce's rigorous approach to artificial intelligence is the real-time jet engine health monitoring service, launched in 1999. Around 3000 engines with dozens of sensors are in the sky at any one time, generating a volume and multiple dimensions of data that are impossible for humans to process. Large amounts of multidimensional information must be processed in near real-time to ensure passenger safety, minimise disruptions, maximise availability and secure revenue. Anyone who has flown long-haul in the last 20 years will probably have had this process going on invisibly in the Rolls-Royce Trent engines next to them on the aircraft wings. This process mirrors the radiotherapy pathway, ensuring the precise and accurate delivery of dose, minimising linear accelerator downtime or treatment interruptions, and ensuring patient throughput. Similar monitoring and feedback of de-identified data from radiotherapy clients could also be beneficial to artificial intelligence service providers to better account for variance in their systems.
New applications of artificial intelligence are frequent. Examples in Rolls-Royce alone include a new method for accelerating borescope inspections of jet engines. Proposed pipeline projects include scanning, recognising and categorising features on high value components. This current, manual quality check could be automated, leading to some ethical considerations but also trustworthiness challenges. Currently, European airworthiness regulators have not approved the use of artificial intelligence in safety-critical applications, so Rolls-Royce's critical artificial intelligence activities are aligned to a proposed regulation roadmap for the ethical and safe advancement of artificial intelligence in aviation [
]. Similar monitoring and quality assurance would be useful for continuous quality monitoring for both linear accelerators and critical software systems for treatment planning and radiotherapy department workflow.
Aletheia is Rolls-Royce's proposal for deploying artificial intelligence ethically in all of its business contexts and was put into place in December 2020. Starting with authoritative guidance, including EU ethics guidelines, the Asilomar principles, Rolls-Royce assurance specialists applied their product safety mentality to create a 32-step process for applying and evidencing what was highly abstract guidance into daily industrial contexts.
The process starts prior to the deployment of an artificial intelligence, with the full and transparent consideration of ethical implications of its proposed activities, particularly from a social impact perspective, but also including a safety and bias viewpoint, before moving to a five-layer high-frequency checking system to ensure that an artificial intelligence's decisions are not wrong and can be trusted. These combine expectation bounding, synthetic data exercising, independence, comprehensiveness and data corruption assurance. Similar processes could be applied to decision support tools for treatment decisions, atlas-based auto-segmentation or magnetic resonance imaging for dose calculation [
]. Transparency and robust data are essential to ensuring artificial intelligence systems engender trust among healthcare professionals who, like airline pilots and flight crews, are intermediary clients trying to ensure each patient is treated responsibly in ways that minimise error, bias and the potential for harm [
Figure 1 shows that in addition to ‘trust and accuracy’ and ‘governance’, the framework also includes ‘social impact’, e.g. the impact of loss of skills and need for retraining: this aspect is also highly relevant to healthcare in general, and radiotherapy in particular [
]. The implementation of artificial intelligence in radiotherapy will have an impact on the tasks performed by doctors, dosimetrists, physicists and radiographers. Artificial intelligence systems may take over certain aspects of clinical decision making, contouring and quality assurance. With deeper integration of artificial intelligence in radiotherapy, oncology professionals will have to emphasise certain aspects of training and may need to learn new, complementary skills that evolve as artificial intelligence replaces some human tasks.
The extensive peer review of Aletheia, with subject matter experts in Big Tech, academia, pharmaceuticals and governments, indicated that it was probably the first framework to bridge the gap between the ‘what’ and the ‘how’ and include trustworthiness. Although there are many other frameworks now in the artificial intelligence stratosphere, they are for the most part, guidance-led documents that do not provide a method to operationalise ethical practice like Aletheia can.
Aletheia assures trustworthiness of the artificial intelligence by focussing on the inputs and outputs on either side of algorithms, not the encoding of the algorithms themselves (which are subject to quality assurance during development). This process allows the continued assurance of the algorithmic outputs, throughout its life, independent of the ‘black box problem’. This also makes it relatively fast to implement and, although originally created to solve an internal Rolls-Royce challenge, applicable in varying applications beyond its safety-critical, industrial manufacturing artificial intelligence origins.
Recognising the need for the responsible use of artificial intelligence in society generally, and the potential for multi-applicability of ethical practice, in 2020 Rolls-Royce openly published Aletheia. Early collaboration to test its flexibility included use cases in the music industry and education, with The Institute for Ethical AI in Education crediting it as informing and influencing the structure of The Ethical AI Framework for AI in Education (UK) published in 2021 [
]. The technical nature of radiotherapy lends itself well to adopting the Rolls-Royce model, but Aletheia's emphasis on ethics may align well with the humanistic core of healthcare.
The Framework continues to evolve, and a second updated version will be published in December 2021, including a module with a step-by-step process for assessing, identifying and mitigating bias risk in artificial intelligence requirements, algorithms and datasets that are used in the development and use of artificial intelligence. Oncology professionals are already finding ways to implement artificial intelligence into healthcare, but can we do it better? We have an opportunity to learn from other industries how to maximise the therapeutic ratio between the benefits and harms of artificial intelligence. For both patients and professionals, it would be prudent to take advantage of any opportunities for collaboration that may make the transition better.
Conflicts of interest
R. Hallows reports a relationship with Rolls-Royce plc that includes: employment.