King's College London · FACULTY OF MANAGEMENT

6QQMN331 Chap.6 Sustainability and Human-Centred AI

- one subject, every graph, every model, every mark
6 Chapters3-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 6 of 8 · 6QQMN331

Sustainability and Human-Centred AI

Evaluate both sides of sustainable AI

This chapter compares AI for sustainability with the sustainability of AI. Organisational benefits must be traced into a real workflow, while compute, energy, water, hardware, data and labour remain part of the decision.

The analysis stays evidence-led and avoids invented impact numbers.

Make human control meaningful

Human-centred design requires understandable roles, appropriate trust, authority to override and organisational responsibility.

Original cases turn these principles into a baseline, pilot, exception route and net-value review.

Use a common objective for comparison

Options can be compared only when they address the same service outcome. A large model, a smaller method and process redesign may produce different benefits and costs, so the analysis first defines the operational decision and baseline.

It then asks how the proposed system changes behaviour, which resource drivers matter and what evidence is missing. This approach permits a defensible recommendation without pretending that unsupported impact numbers are known.

Look beyond the first efficiency gain

An intervention may reduce resource use per task while lowering effort or cost enough to increase total demand.

The chapter uses this rebound possibility to test the boundary of the benefit claim. Human-centred analysis adds the people who operate, experience and challenge the system. Effective control requires understandable information, real override authority and organisational response to override patterns.

The final net-value judgement therefore includes service performance, system costs, affected groups and the evidence that would justify continuing or withdrawing the use.

In this chapter

What this chapter covers

  • 01

    AI for Sustainability

  • 02

    Sustainability of AI

  • 03

    Baseline

  • 04

    Net Value

  • 05

    Human Control

  • 06

    Calibrated Trust

Worked example · free

Redesign a sustainability proposal

Q [8 marks]. An organisation proposes a large model for energy scheduling but provides no baseline, comparison or human override. Repair the proposal. The marks shown are an AskSia study aid, not an official University marking scheme.
Dual sustainability test.Dual sustainability test. Organisational benefits from AI must be compared with the environmental and social costs of the system itself. AI for sustainabilityorganisational benefitSustainability of AIsystem costshuman control sets the acceptable trade-off
Dual sustainability test. Organisational benefits from AI must be compared with the environmental and social costs of the system itself.
  • 2Define the baseline and operational decision.
  • 3Compare benefit with system cost and alternatives.
  • 3Set control, evidence and expansion conditions.
Measure whether scheduling changes reduce energy while maintaining service, compare the large model with a smaller method and request evidence about operational resource use. Facilities staff retain override authority. Expansion depends on measured net improvement, stable service and evidence that system costs do not erase the benefit.
Sia tip — Evaluate each option against the same service objective before comparing resource use.
Glossary

Key terms

Sustainable AI
Sustainable AI evaluates both AI-enabled benefits and the environmental or social costs of the system.
Net Value
Net Value is the justified benefit remaining after relevant system costs and trade-offs are considered.
Human Control
Human Control requires information, authority and an actionable route to change or stop the system.
Rebound Effect
A Rebound Effect occurs when efficiency lowers cost or effort and stimulates demand that erodes the gain.
FAQ

Sustainability and Human-Centred AI FAQ

What are the two sides of sustainable AI?

AI for sustainability concerns organisational or social benefits enabled by AI. The sustainability of AI concerns the environmental and social costs of building and operating the system. A defensible recommendation compares both sides against the same outcome and affected groups.

How can sustainability be assessed without invented numbers?

Define the baseline, identify the main benefit and cost drivers, request evidence from relevant teams or suppliers and compare alternatives that can meet the same task. State uncertainty and use a bounded pilot rather than assigning unsupported carbon or resource estimates.

What makes human control meaningful?

People need to understand the system's role, recognise unreliable conditions, possess authority to override or stop it and avoid penalties for justified challenge. Managers must review override evidence and remain responsible for the workflow rather than treating a button as sufficient oversight.

Why compare against the same service objective?

A resource comparison is misleading when options deliver different outcomes. Define the required service, quality and reliability first, then compare ways to achieve it. This reveals whether a larger model creates additional value or merely consumes more resources for a result that a smaller method or process change could supply.

How does a rebound effect change the analysis?

Efficiency can lower the effort or cost of a task and stimulate more use, reducing the expected total saving. Identify the behaviour that might change and include it in monitoring. The possibility does not cancel an efficiency gain, but it prevents a per-task improvement from being reported automatically as an overall reduction.

What should a sustainability pilot measure?

Measure the intended operational benefit, relevant system resource drivers, service quality, exceptions and human workload against a stated baseline. Include a simpler alternative where possible. The review should decide whether net improvement is supported, which uncertainties remain and whether the use should expand, narrow or stop.

How should uncertain impact evidence be reported?

Separate what has been measured from what remains an evidence request. Name the relevant resource or social cost, the party able to provide data and the decision that depends on it. Use conditional recommendations and a bounded test instead of inventing precision. Uncertainty should narrow the claim while preserving a clear next action and accountable owner.

Record when the missing evidence will be reviewed and what result would change the proposed scope.

Study strategy

Assessment move

For each sustainability claim, write the baseline, AI-enabled decision change, intended benefit, system costs, affected groups and human authority. Compare a simpler option and predict one rebound effect. Close with a limited test and evidence that would justify expansion, revision or withdrawal. Check that every option serves the same quality and reliability objective.

Separate measured effects from requested evidence, then describe how operators recognise an unreliable condition, use override authority and return the reason to organisational review.

Working through Sustainability and Human-Centred AI in 6QQMN331? Sia is AskSia’s AI Management tutor — ask any 6QQMN331 Sustainability and Human-Centred AI question and get a clear, step-by-step explanation grounded in how 6QQMN331 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 18 of your subjects at King's College London, plus 1,000+ more Bibles across the AskSia library.
Sia - your 6QQMN331 tutor, unlimited, worked the way the exam marks it
The full 3-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works