GCS Framework for Ethical Innovation
Continuing Professional Development (CPD) points: 2
Table of contents
- Introduction
- Explainer: our core values
- Overview
- Framework in action: Case study 1
- Framework in action: Case study 2
- Step 1 – UK Legal Frameworks
- Step 2 – Propriety Guidance
- Step 2 (continued) – GCS Propriety and Ethics guidance
- Step 3
- Additional considerations
- What should you do next?
Introduction
The purpose of this framework is to support GCS members to navigate, understand, and apply our values, to guide the use of emerging technologies and applications of data.
The framework sets out the safeguards we will adopt to assess innovative new approaches, and aims to provide government communicators with the confidence that their approach is in line with our ethical standards.
New ideas will rarely be 100% ethical or unethical. This framework is designed to help all members identify what they should be considering from the outset of their innovative work, and guide thinking using our values, to make the best decisions possible.
The framework will be regularly reviewed and updated by the central Government Communication Service (GCS) team, to keep pace with emerging opportunities, risks, and official guidance.
Explainer: our core values
British values such as Democracy, the Rule of Law, and Human Rights are the foundations of our society, and are therefore important principles we should consider. For example, Democratic Accountability means we should always act in a transparent manner, and be able to fully explain or justify an approach or how a decision was made to our Minister(s). Considering Human Rights includes any impact on the fundamental entitlements and freedoms for all. For example, not impinging on individual liberties through adoption of new technologies. These values are especially important when it comes to our use of generative AI.
These are then built upon by the Civil Service Code, which sets out our core values of integrity, honesty, objectivity, and impartiality. For example, Honesty is essential for maintaining trust in democratic institutions, and we strengthen this through communicating our actions transparently.
Next, the GCS code of conduct governs how the government communicates with the public. For example, government communications have to be accurate and truthful, ensuring that all information presented to the public is reliable and factually accurate. Therefore, this is particularly important for our use of innovative technologies like generative AI. We need to do so in a way that upholds accuracy, inclusivity and mitigates biases.
To summarise:
British Values:
- Democratic Accountability
- Human Rights
- UK Legal Frameworks
Civil Service Code:
- Integrity
- Honesty
- Impartiality
- Objectivity
The Government Publicity Conventions:
- Not Party Political
- Objective and Explanatory
- Economic and Appropriate
- Relevant
Overview
This framework is comprised of three steps designed to be worked through in turn:
- Step 1 – Considering legal obligations
- Step 2 – Considering propriety guidance
- Step 3 – Reviewing topic specific policies and frameworks
If you are planning a complex innovation project, you can use the guide linked below to support with this.
This provides best practice guidance on how to map stakeholders which may be impacted by your work. It also includes a template assessment grid, to help record your findings as you review the ethical considerations of your planned activity: 2024-01 GCS Framework for Ethical Innovation – Complex cases assessment sheet.
Framework in action: Case study 1
Opportunity
AI Tools to enable the automated hiding of comments below an advertiser’s posts on social media platforms.
Potential benefits for government communications
- Time saved moderating comment sections.
- Reducing the exposure to harmful language for those engaging with government social content.
- Minimising the impact of harmful language on campaign messaging.
Issues identified through applying this framework
- The technology may transfer data across different international regions, raising Data Protection risks with inadvertent sharing of personal data.
- Though hiding comments made by the public, use of the technology may impact upon principles of the Human Rights Act, specifically, impinging upon individual’s Freedom of Expression.
Result
- In the form offered, this technology is not currently suitable for use in government communications. Further research is required to identify whether there are specific applications which would justify the additional spend on this type of tool, over simply turning off comments on posts.
Framework in action: Case study 2
Opportunity
- AI powered tools for assessing diversity in digital communications content, in order to identify opportunities for improvement.
Potential benefits for government communications
- Support efforts to ensure government communications are representative of the audiences we are trying to reach and represent.
Issues identified through applying this framework
- For the tool offered, the underlying AI technology is trained on demographic data from the United States, which may not be fully representative of the unique demographic landscape of the UK.
Result
The technology is not yet mature enough to avoid perpetuating bias in its results within a UK context, and so not yet suitable for use in government communications.
Step 1 – UK Legal Frameworks
Your planned activity or approach be compliant with existing laws and regulations. For example, this might include:
UK Data Protection Act
- Data Processing: Personal data should only be collected and used where there is a clear legal justification, and that this can be communicated clearly to data subjects as necessary.
- Impact Assessments (DPIAs): Data Protection Impact assessments should be carried out for any new or large scale data processing, or use of sensitive data to identify and mitigate risks.
- Information Commissioner’s Office (ICO): Any data breaches must be reported to the ICO within 72 hours.
UK GDPR
- Data Processing: Individuals must consent to their data being processed; you should obtain explicit opt-in consent before collecting or sharing information.
- Data Rights/Transparency: Individuals have the right to transparency on who accesses their data, why it was collected, and how it is used; you must be able to provide disclosure when required.
- Data Transfers: Consider if any data will be transferred outside of the UK, and if you have the appropriate safeguards in place to mitigate risks associated with this.
Equalities Act
- Inclusivity/Representation: Activity should respect inclusivity and equal representation, and we should not perpetuate biases or stereotypes in our work.
- Accessibility: We should consider how our work can be made as accessible as possible to people with disabilities. This means providing information in accessible formats, such as large print, and ensuring digital content meets accessibility standards wherever possible.
Intellectual Property Act
- Trademarks: Trademark protection would need to be considered in the use of any official insignia, logos, or service marks.
- Patents: Any patented processes or technologies being used in the creation or dissemination of communications would need to comply with patent laws.
- Copyrights: The UK government often has a permitted right to use copyrighted material for official purposes under ‘Crown Use’ provisions.
Please note: this is not an exhaustive list. There may be other specific laws, acts, or regulations to consider which may be relevant to your organisation.
Step 2: Propriety Guidance
The Civil Service Code
Will you be able to uphold the Civil Service Code through your planned activity?
For example, but not exclusively:
- Will your planned activity prevent you from making decisions that are solely based on evidence, and without favouritism?
- Will you be able to act in a transparent manner about your planned activity? For example, will you be able to fully explain or justify an approach or how a decision was made to your Minister?
- Will your planned activity uphold political neutrality, avoiding involvement in partisan politics and maintaining a focus on serving the government of the day?
- Will you be able to explain your planned activity? For example, ensuring the relevant people can access, interpret and understand your activity?
Remember to consider:
- Integrity – Putting the obligations of public service above personal interests
- Honesty – Being truthful and open
- Impartiality – Basing advice and decisions on rigorous analysis of the evidence
- Objectivity – Acting solely according to the merits of the case and serving governments of different political parties equally well.
Step 2 (continued): GCS Propriety and Ethics guidance
GCS Propriety and Ethics (P&E) guidance highlights the importance of applying the Government Publicity Conventions in our work.
Will you be able to uphold these through your planned activity?
Not Party Political
The communication should not be – or liable to be – misrepresented as being party political.
It must not justify or defend policies in party political terms, use political slogans, or advocate policies of a particular political party or directly attack policies and opinions of opposition parties and/or groups.
Will your activity uphold the principle of political impartiality, ensuring that communications clearly explain the Government of the day policies’ in a balanced and objective way, whilst serving the best interests of the public?
Relevant
The communication should be relevant to government responsibilities.
Matters dealt with by government publicity should be ones in which the Government has direct and substantial responsibilities. The Government should explain and justify its policies and decisions, and, when necessary, inform, advise, alert or warn the public.
Will any content produced through your planned activity be demonstrably relevant to government responsibilities, and not stray into areas which are party political, or which cannot be linked back to core government responsibilities?
Economic and Appropriate
The communication should be conducted in an economic and appropriate way and should be able to justify the costs as expenditure of public funds.
The use of public funds should be appropriate and effective in their application. The resources should be justifiable on value-for-money grounds.
Does any expenditure represent value for money for taxpayers by delivering an outcome which is a public good in the most economic way?
Objective and Explanatory
The communication should be objective and explanatory, not biased or polemical.
Government communications should be objective as possible, personalisation of issues and biases should be avoided and any communications should always come from a perspective of informing the public.
Through your planned activity, will you be able to be accurate and truthful, ensuring that all information presented to the public is reliable and factually accurate?
Please note: the GCS P&E guidance covers additional areas which may not be covered here, but relevant to your planned activity.
Step 3
Assess your planned activity against existing GCS policies and guidance on specific topics
Will you be able to adhere to, and mitigate any conflicts with, those that are applicable to your planned activity?
Activity 1: I’m planning to use a new channel or platform
Refer to the GCS SAFE Framework:
- Safety and suitability
- Ads context
- Freedom of speech
- Ethics and enforcement
Activity 2: I’m planning novel use of first-party data
Refer to the GCS Innovating with Impact Strategy.
3 steps to guide use of our first-party data:
- Map and discover
- Secure consent and communicate
- Minimise and innovate
Activity 3: I’m planning use of generative AI
Refer to the GCS Generative AI policy.
- How we will use generative AI.
- How we will not use generative AI.
- How we will act transparently in our use.
Additional considerations
For complex innovation projects, please refer to the guidance below:
Activity: I’m planning the development of a new AI tool
Refer to the Generative AI framework for HMG.
- Security of AI tools: sensitive information should be safely handled and inline with data protection legislation.
- Meaningful human control: Processes should be in place to ensure that AI outputs are checked and validated by qualified personnel.
- Lifecycle management: From selection to decommissioning, the lifecycle of AI tools must be managed carefully.
- Choosing the right tool: Selecting the most suitable AI technology for the task at hand is vital to ensure effectiveness and cost-efficiency.
- Openness and collaboration: Engaging widely with other departments and stakeholders is encouraged to foster transparency and innovation.
- Integration with commercial practices: Collaboration with commercial partners should maintain high standards of transparency and ethical practice.
- Skills and expertise: Building and maintaining AI tools requires specific skills, which should be developed within an organisations personnel.
- Compliance and assurance: AI use should align with existing organisational policies and governance structures.
Activity: I’m planning novel use of first-party data
Refer to the HMG Data Ethics Framework.
- Value and public benefit: Data use should be clearly linked to public good, ensuring that the benefits of any data initiative are maximised for the public while respecting the rights and interests of individuals.
- Clarity and transparency: Activities involving data should be understandable and transparent to the public. The reasons for decisions and the methodologies used should be open and accessible.
- Accountability and responsibility: Those working with data should remain accountable for their actions. This involves clear documentation of decisions and ensuring responsibility for data use is always attributed.
- Ethical practices and governance: Data use should comply with ethical standards and legal requirements, with clear governance structures in place to guide decision-making and oversee the responsible use of data.
- Equity and fairness: Individuals and communities should be treated fairly, with measures to avoid and address biases, discrimination, and the unintended consequences of data use. Data must be used in ways that strive to be inclusive and equitable.
Activity: I’m planning to develop or adopt a new data model or digital tool in my work
Refer to: Responsible Tech Adoption Unit (RTA) Principles for Responsible Innovation
- Stay updated on evolving legal and ethical requirements and adapt technology accordingly.
- Regularly conduct audits and assessments to verify compliance where this is legally required, and identify areas for improvement.
- Collect feedback from users, stakeholders, and relevant experts to identify areas for improvement and address ethical concerns.
- Continuously adapt and enhance the technology and its supporting processes to align with evolving ethical standards and societal expectations.
- Establish mechanisms for reporting and addressing ethical concerns or dilemmas related to the technology’s use.
- Provide accessible information about data handling practices, algorithms, and decision-making processes to promote transparency and informed decision making?
- Clearly communicate to users and stakeholders how the technology works, its intended purpose, and any potential risks or limitations, if needed.
- Take responsibility for any errors or biases in the technology and establish mechanisms for redress and rectification.
Refer to: Algorithmic Transparency Recording Standard (ATRS)
- Transparency and accessibility: Information about the use of algorithmic tools must be openly published in an understandable and easily accessible way via a algorithmic transparency report. This should include details about how these tools support any decision-making processes.
- Comprehensive information: This detailed documentation should include descriptions of the tool’s design, operation, the decision-making processes it influences, the data it uses, any associated risks, and measures taken to mitigate these risks.
- Clear assignment of responsibilities: Where appropriate, organisations are required to designate specific roles such as an Algorithmic Transparency Lead, who oversees the completion of transparency records. This role involves coordinating with various departments to gather necessary information and ensure compliance with the standard.
- Inclusion of all relevant stakeholders: The process of completing a transparency record involves contributions from various sectors of an organisation, including operational teams, data science teams, legal, and ethics departments, ensuring a holistic representation of how algorithmic tools are used.
What should you do next?
If you have identified that there are barriers preventing you from upholding either our legal obligations, propriety guidance, or best practices and policies, then you should:
- Pause activity for now.
- Carefully review the risks or barriers you have identified, and consider where further scrutiny or attention is required to uphold or abide by these before proceeding.
- Engage with domain experts in the area where further attention is required, to support you to strengthen your approach, or identify other innovative means to achieve your objectives.
- If your use case involves the application of AI, our AI Policy on the GCS website outlines additional relevant guidance you can use to help you. You should also review the CDDO Framework on generative AI use.
- Once ready, reassess your planned activity using this tool.
GCS Connect can help you to identify colleagues across the profession who may be able to support.
If you have reviewed all relevant steps for your planned activity, and identified the risks to be low, then it’s now over to you for the final step – to test and scale!
We encourage you to continuously evaluate your activity against this framework, and continually engage with experts and peers to help you.
We also want to hear from you – please contact the central GCS team if:
- You have feedback or thoughts about how we can improve this ethical framework
- You need more guidance on how to you can identify, test, and scale innovative ideas
- You want to know more about best practice around insight and evaluation
- You want to know more about best practice in data modelling and visualisation
- You want to know how to lead effective campaigns using data
The GCS Applied Data & Insight team is able to advise on best practice and approaches (AI.GCS@cabinetoffice.gov.uk).