Principles into Practice 2 2 2 2
What do we mean by "assessing for the full range of effects"?
Filed under:
Human Centricity
Responsibility
Understanding
Bias and Harm Mitigation
Reliability

AI systems must account for a spectrum of effects, both direct and indirect, positive and negative, across all operational contexts. 
 
Positive effects can include: 
  • Enhanced decision-making through data-driven insights.
  • Reduced risk to military personnel/non-combatants via automation of hazardous tasks.
  • Improved military precision, potentially lowering collateral damage.
  • Strengthened operational efficiency, benefitting national and global security.
  • Cost reduction

Negative effects can include:
  • Risk of over-reliance on AI, undermining human agency.
  • Potential for unintended consequences, such as discrimination or biased outcomes.
  • Mental and emotional impacts on operators, such as stress or reduced autonomy.
  • Ethical concerns regarding privacy, cognitive liberty, and moral responsibility.

The distinction between direct and indirect effects is often used across various fields, such as science, economics, law, and military strategy, to describe how actions or interventions lead to consequences. Direct effects refer to the immediate and primary outcomes of an action, occurring as a direct result of the cause without the involvement of intermediary factors. These effects are typically predictable and have a clear, observable causal link to the source. For example, in physics, pushing a ball causes it to roll; in law, enforcing a smoking ban directly reduces smoking rates in public spaces; and in military strategy, a bomb destroying a target structure is a direct effect of the attack.

Indirect effects, on the other hand, are the secondary or tertiary outcomes that arise as a consequence of the direct effects, often mediated by additional factors or processes. These effects are broader, may take time to manifest, and are often less predictable. For instance, in economics, raising interest rates directly increases borrowing costs, but the indirect effect may include slower economic growth. Similarly, in law, a smoking ban indirectly reduces healthcare costs over time by lowering smoking-related illnesses. In a military context, destroying a power plant might cause economic disruption and civilian displacement as indirect effects of the initial action.

The key differences between direct and indirect effects lie in their causation, predictability, timeframe, and scope. Direct effects are immediate and clearly linked to the cause, often occurring in the short term, with a limited scope confined to the initial action. In contrast, indirect effects are mediated by other factors, take longer to manifest, and often involve broader, cascading outcomes. Understanding this distinction is crucial for analysing results and planning effectively, particularly in policy-making, strategy, and other fields, as it helps anticipate unintended consequences and manage the broader implications of actions.

Disclaimer

This tool has been created in collaboration with Dstl as part of an AI Research project. The intent is for this tool to help generate discussion between project teams that are involved in the development of AI tools and techniques within MOD. It is hoped that this will result in an increased awareness of the MOD’s AI ethical principles (as set out in the Ambitious, Safe and Responsible policy paper) and ensure that these are considered and discussed at the earliest stages of a project’s lifecycle and throughout. This tool has not been designed to be used outside of this context. 
The use of this information does not negate the need for an ethical risk assessment, or other processes set out in the Dependable AI JSP 936 part 1, the MODs’ policy on responsible AI use and development. This training tool has been published to encourage more discussion and awareness of AI ethics across MOD science and technology and development teams within academia and industry and demonstrates our commitment to the practical implementation of our AI ethics principles.