Ethics + 2027 Case Study + final Paper 1 synthesis
Retrieve all official terminology, build balanced ethical arguments and answer the four Visionary Studios challenges.
Final synthesis
- discuss accountability, algorithmic fairness, bias, consent, environmental impact, privacy, security, societal impact and transparency;
- discuss misinformation, online harassment, anonymity and privacy in online communication;
- reassess ethics as quantum computing, augmented reality, virtual reality and pervasive AI affect society, individual rights, privacy and equity;
- retrieve all 26 official case-study terms;
- answer all four official case-study challenges with mechanism, evidence, trade-offs and judgement.
Official Case-Study Terminology Deck
Attempt each card first. If a term is new, flip and study it, then answer it again without looking. Mark only secure recall as “Know it”.
Machine-learning ethics
Applicant-ranking system: choose four ethical lenses
A company uses an ML system to rank applicants for interviews. Choose four ethical lenses from the reference list and discuss how each could affect applicants or the company. Your four must include algorithmic fairness/bias and accountability. Finish with one safeguard or governance action.
The organization remains accountable. Training data or features can create unfair group outcomes; personal data requires appropriate consent and protection; meaningful transparency and appeal routes are needed; automation can scale consistency but also historical inequality. Resource use creates environmental impact. A defensible deployment needs audits, oversight, security, governance and continuing review.
Online communication and changing technology
Online communication
An online platform uses ML to recommend and moderate posts. Choose two issues from misinformation, online harassment, anonymity/accountability, and privacy. For each, explain how algorithmic recommendation or moderation could reduce the problem or make it worse.
Risks include misinformation amplification, bias, online harassment exposure, privacy loss and tensions between anonymity and accountability. Benefits such as moderation and relevance must be weighed against these harms.
Reassess ethics as technology changes
Choose two of these emerging technologies: quantum computing, augmented reality, virtual reality, pervasive AI. For each chosen technology, explain one way it could change an existing ethical concern involving society, individual rights, privacy or equity. Then explain why guidelines may need updating as capabilities change.
Examples: quantum computing can change assumptions about cryptographic security; AR/VR can collect intimate spatial, biometric or behavioural data; pervasive AI can influence employment, access to services and decisions at scale. These shifts can create new privacy, rights or equity consequences, so guidance written for older capabilities may no longer address the real harms or safeguards needed.
Diffusion: quality vs compute
Managing iterative denoising
Visionary Studios must manage the computational demands of iterative denoising while still producing photorealistic images. Build a case-specific response that explains the diffusion mechanism, uses one relevant fact/example from your research pack, evaluates the resource trade-off and reaches a recommendation.
Diffusion repeatedly applies a learned CNN denoiser from a noisy starting point under a DDPM-style process. Repetition drives computation, memory traffic and latency. Accelerators plus cloud or HPC can help but add cost, energy use and infrastructure dependence. Visionary Studios should judge whether photorealistic quality justifies these costs.
Dataset intellectual property + bias mitigation
Large training datasets
Visionary Studios must use large training datasets without ignoring intellectual-property or bias risks. Build a response that explains the data-governance problem, uses one researched example/fact, evaluates safeguards and reaches a justified recommendation.
Large datasets broaden learned patterns but increase provenance, intellectual-property and bias risks. Visionary Studios should document sources and rights, resolve material that cannot lawfully be used, audit representation and output bias, apply mitigation and disclose AI use appropriately. Scale should not override lawful, representative and traceable curation.
Generator/discriminator balance
GAN training stability
Visionary Studios must balance generator and discriminator learning so a GAN produces realistic and varied images. Explain the adversarial mechanism and the risk of imbalance or mode collapse. Then use one mitigation idea from your previously researched source to evaluate what the studio should try. If your research pack does not yet contain evidence, mark that as a gap rather than inventing a source.
The generator maps noise vectors to synthetic images and learns to fool the discriminator; the discriminator improves at separating real from fake. Severe imbalance can weaken useful learning signals and destabilize training. Mode collapse reduces diversity. A recommendation should use researched mitigation evidence and judge sharpness, diversity, consistency, reliability and compute together.
Hybrid model evaluation
Balance model families
Visionary Studios is considering a hybrid that combines elements of diffusion models, GANs, VAEs and flow-based models. Compare the model families only as deeply as required, use the official evaluation factors, and decide whether a hybrid is justified for the studio's needs rather than assuming that more components are automatically better.
A hybrid can combine diffusion quality, GAN sharpness, VAE latent-space flexibility and flow-based reversibility or traceability. But architecture, training and deployment complexity rises. Compare output quality, computational efficiency, training stability, flexibility and scalability, and consistency rather than assuming more components are better.
Consistency and disclosure
Character consistency vs style adherence
Distinguish them.
Character consistency preserves recurring appearance across images; embedding-based approaches can help. Style adherence means conditional outputs follow the intended artistic style or theme specified by prompts.
AI disclosure
Why can transparency matter even when output quality is high?
Clients and audiences may need clarity about AI involvement, provenance, authorship and accountability. Technical quality does not resolve trust or ethical and legal transparency.
Fundamentals + preprocessing
Paper 1 checkpoint A
1. Distinguish supervised, unsupervised and reinforcement learning. 2. Why GPU for CNN? 3. Give two cleaning techniques. 4. Explain one curse-of-dimensionality issue.
Supervised uses labels; unsupervised discovers unlabeled structure; RL learns policies from reward interaction. GPU provides parallel tensor throughput. Cleaning examples include duplicates, errors, missing, misformatted or irrelevant data. High dimensions can cause sparsity, overfitting, weak distance metrics or compute and memory growth.
Approaches + neural networks
Paper 1 checkpoint B
1. Interpret regression slope. 2. Explain K-NN. 3. Define precision vs recall. 4. Explain exploration vs exploitation. 5. Outline a perceptron and basic CNN.
Slope is predicted response change per predictor unit. K-NN derives labels from nearby labelled examples. Precision is correctness among predicted positives; recall is captured actual positives. Exploration learns uncertain actions; exploitation uses current knowledge. Perceptron: weighted inputs plus bias → activation → output. CNN: input → convolution, activation and pooling → fully connected → output.
Six-move answer plan
Choose one official challenge
Finish the learning cycle
Case study / exam preparation
Finalize the research pack around the four official challenges with credible evidence attached to claims.
Paper 1 — short practice
Redo only identified A4 gaps, then complete one timed Paper 1 ML set and one timed case-study response.
Case-study checkpoint
Verify all 26 official terminology items and all four official challenges can be explained without notes.