A Level H1 General Paper Practice Paper 4

TuitionGoWhere Practice Paper - General Paper H1 A-Level

Answer Key and Marking Scheme

Subject: General Paper H1
Paper: Practice Paper (Comprehension)
Version: 4 of 5
Total Marks: 50

Section A: Comprehension (35 marks)

Question 1 (2 marks)

Question: According to the passage, what is "algorithmic governance"? Use your own words as far as possible.

Answer: Algorithmic governance is the practice of giving decision-making power to computer systems that analyse large amounts of data to make decisions that were previously made by humans.

Marking Scheme:

• 1 mark for identifying that it involves delegating decisions to computational/algorithms.
• 1 mark for noting that these systems process data to make decisions formerly made by humans.
• Common mistake: Lifting the phrase "delegation of decision-making authority to computational systems" directly without paraphrasing. Students must use their own words.

Question 2 (2 marks)

Question: Explain the author's use of the phrase "a quiet revolution" in line 1. Use your own words as far as possible.

Answer: The phrase "a quiet revolution" suggests that the shift towards algorithmic decision-making is a profound and far-reaching change, but one that is happening gradually and without widespread public awareness or protest. The word "quiet" implies that this transformation is not dramatic or visible like a traditional revolution, yet its effects are significant.

Marking Scheme:

• 1 mark for explaining that it refers to a significant but unnoticed/gradual change.
• 1 mark for noting the contrast between the magnitude of the change ("revolution") and its lack of visibility ("quiet").
• Common mistake: Only explaining "revolution" without addressing the significance of "quiet."

Question 3 (2 marks)

Question: What does the word "opacity" (line 28) suggest about algorithmic systems? Use your own words as far as possible.

Answer: "Opacity" suggests that algorithmic systems are difficult or impossible to understand — their decision-making processes are hidden or unclear, even to the people who created them. This means individuals affected by algorithmic decisions cannot see or understand how those decisions were reached.

Marking Scheme:

• 1 mark for identifying that it means the systems are unclear, hidden, or impossible to understand.
• 1 mark for linking this to the inability of people to see or challenge how decisions are made.
• Common mistake: Simply defining "opacity" as "not transparent" without explaining what this means in context.

Question 4 (2 marks)

Question: According to lines 8–14, what are two advantages of algorithmic systems over human decision-makers? Use your own words as far as possible.

Answer:

1. Algorithms apply the same criteria consistently to every case, eliminating the unpredictable biases that affect human decisions (such as anchoring or confirmation bias).
2. Algorithms can process vast numbers of applications extremely quickly, far faster than any human committee could manage.

Marking Scheme:

• 1 mark for each advantage, up to 2 marks.
• Acceptable answers include: consistency/uniformity, elimination of human cognitive bias, speed/efficiency, ability to handle large volumes.
• Common mistake: Stating "they are objective" without explaining what this means (i.e., consistent application of criteria).

Question 5 (3 marks)

Question: Explain the author's use of the phrase "the coded gaze" (line 39). What does this phrase reveal about algorithmic systems? Use your own words as far as possible.

Answer: "The coded gaze" refers to the embedded assumptions and prejudices that are built into algorithmic systems through their programming and training data. The word "gaze" suggests a perspective or way of looking at the world, while "coded" indicates that this perspective is embedded in the software. The phrase reveals that algorithms are not neutral — they carry the biases of their creators and the data they are trained on, and these biases can perpetuate inequality while appearing mathematically objective.

Marking Scheme:

• 1 mark for explaining that it refers to embedded assumptions/bias within algorithms.
• 1 mark for noting that "gaze" implies a perspective or viewpoint.
• 1 mark for explaining that this reveals algorithms are not truly neutral — they reflect existing prejudices.
• Common mistake: Only explaining the literal meaning without connecting it to the author's argument about bias.

Question 6 (2 marks)

Question: What is meant by the "accountability gap" (line 47)? Use your own words as far as possible.

Answer: The "accountability gap" refers to the situation where harmful decisions made by algorithms cannot be attributed to any single responsible party. Because multiple actors are involved (data scientists, managers, organisations), blame is diffused, and no one person or entity takes responsibility for the negative outcomes.

Marking Scheme:

• 1 mark for identifying that it refers to a lack of clear responsibility.
• 1 mark for explaining that blame is spread across multiple parties, making it difficult to hold anyone accountable.
• Common mistake: Confusing "accountability gap" with "transparency gap" — the issue is about responsibility, not understanding.

Question 7 (2 marks)

Question: According to lines 50–57, what two different regulatory approaches to algorithmic governance are mentioned? Use your own words as far as possible.

Answer:

1. The European Union has adopted a legislative approach, granting individuals the right to meaningful explanation when subject to automated decisions (through the GDPR).
2. Singapore has adopted a more flexible, voluntary approach, issuing governance frameworks and encouraging industry self-regulation rather than imposing strict rules.

Marking Scheme:

• 1 mark for each approach, up to 2 marks.
• Acceptable answers: EU's legal/rights-based approach vs. Singapore's voluntary/pragmatic approach; or specific examples such as GDPR's explanation right vs. Singapore's self-regulation model.
• Common mistake: Only naming the countries without explaining the different approaches.

Question 8 (2 marks)

Question: What does the word "legible" (line 66) mean in the context of the passage? Use your own words as far as possible.

Answer: In this context, "legible" means understandable or transparent — that the workings and decision-making processes of algorithmic systems should be clear and comprehensible to the people affected by them, rather than hidden or opaque.

Marking Scheme:

• 1 mark for identifying the meaning as "understandable" or "transparent."
• 1 mark for connecting it to the context of algorithmic systems being open to scrutiny.
• Common mistake: Defining "legible" only in its literal sense (able to be read) without connecting it to the passage's argument about transparency.

Question 9 (3 marks)

Question: In your own words, explain why algorithms can amplify bias even though they do not "intend to discriminate" (lines 37–38). Use your own words as far as possible.

Answer: Algorithms learn from historical data, and this data often contains patterns of past discrimination or prejudice. When an algorithm is trained on such data, it identifies and replicates these statistical patterns, effectively automating and potentially intensifying existing biases. Because the algorithm simply processes correlations in the data without any awareness or intent, it can produce discriminatory outcomes while appearing mathematically neutral.

Marking Scheme:

• 1 mark for explaining that algorithms learn from historical data.
• 1 mark for noting that historical data contains past prejudices/biases.
• 1 mark for explaining that algorithms replicate these patterns without intent, making the bias systemic rather than deliberate.
• Common mistake: Stating that algorithms "are biased" without explaining the mechanism (learning from biased data).

Question 10 (2 marks)

Question: According to the passage, what is the author's overall stance on algorithmic governance? Support your evidence from the text.

Answer: The author adopts a balanced but cautious stance. The author acknowledges the benefits of algorithmic governance (efficiency, consistency, objectivity) but emphasises the serious risks (opacity, bias amplification, accountability gaps). The concluding lines make clear that the author believes algorithmic governance is a powerful tool that must be carefully regulated to serve the public interest.

Marking Scheme:

• 1 mark for identifying the balanced/cautious stance.
• 1 mark for providing textual evidence (e.g., "not inherently good or evil," "the stakes are high," or the call for systems that are "legible, fair, and accountable").
• Common mistake: Claiming the author is entirely for or against algorithmic governance — the passage is deliberately balanced.

Question 11 (2 marks)

Question: Explain the author's use of the word "veneer" (line 40). What does it suggest about the appearance of algorithmic neutrality?

Answer: The word "veneer" suggests a thin, superficial layer that conceals what lies beneath. In this context, it implies that the mathematical neutrality of algorithms is only a surface appearance that masks the underlying biases embedded in the system. The neutrality is not genuine — it is a facade.

Marking Scheme:

• 1 mark for explaining that "veneer" means a superficial or thin covering.
• 1 mark for connecting it to the idea that algorithmic neutrality is a facade hiding real bias.
• Common mistake: Only defining "veneer" without explaining its rhetorical effect in context.

Question 12 (2 marks)

Question: What does the phrase "moral valence" (line 61) mean in the context of the passage? Use your own words as far as possible.

Answer: "Moral valence" refers to the ethical quality or moral character of something — whether it is good or bad. In this context, the author is saying that algorithmic governance is not inherently ethical or unethical; its moral value depends on how it is designed and used.

Marking Scheme:

• 1 mark for identifying that it refers to moral/ethical quality.
• 1 mark for explaining that it depends on context and use, not on the technology itself.
• Common mistake: Not connecting "valence" to the idea of moral value or ethical character.

Question 13 (2 marks)

Question: According to lines 58–63, what is the author's view on whether algorithmic governance is "inherently good or evil"? Use your own words as far as possible.

Answer: The author believes that algorithmic governance is neither inherently good nor inherently evil. It is a tool whose ethical value depends on how it is designed, implemented, and regulated. The moral outcome is determined by human choices about its use, not by the technology itself.

Marking Scheme:

• 1 mark for stating that the author rejects the idea that it is inherently good or evil.
• 1 mark for explaining that its moral value depends on design, deployment, and oversight.
• Common mistake: Oversimplifying the author's view as "it's good" or "it's bad."

Question 14 (3 marks)

Question: How does the author structure the argument in this passage? Identify the key stages of the argument and explain how they build towards the conclusion.

Answer: The author structures the argument in four key stages:

1. Introduction of the concept (lines 1–7): Defines algorithmic governance and establishes its growing importance.
2. Presentation of the case in favour (lines 8–14): Outlines the benefits — objectivity, consistency, and efficiency — to show what proponents argue.
3. Presentation of concerns (lines 15–48): Systematically examines three major risks — opacity, bias amplification, and accountability gaps — each with specific evidence and examples.
4. Conclusion and call to action (lines 58–67): Argues that the technology is a tool whose moral value depends on human oversight, and calls for systems that are legible, fair, and accountable.

The structure moves from definition to balanced analysis to a nuanced conclusion, building a case that the issue is complex and requires careful governance.

Marking Scheme:

• 1 mark for identifying at least two structural stages.
• 1 mark for explaining how the stages build on each other (e.g., balanced presentation leading to nuanced conclusion).
• 1 mark for noting the overall movement from introduction through balanced argument to conclusion.
• Common mistake: Simply listing paragraph summaries without explaining how they function as stages of an argument.

Question 15 (2 marks)

Question: What is the author's purpose in mentioning Singapore specifically (lines 55–57)? Use your own words as far as possible.

Answer: The author mentions Singapore to provide a concrete example of a different regulatory approach — one that prioritises innovation and industry self-regulation over strict legal requirements. This illustrates the diversity of global responses to algorithmic governance and reinforces the author's point that different jurisdictions are grappling with the same challenge in different ways.

Marking Scheme:

• 1 mark for identifying that Singapore represents a pragmatic/voluntary approach.
• 1 mark for explaining that this adds a real-world example showing the range of regulatory responses.
• Common mistake: Only stating that Singapore is mentioned without explaining its rhetorical purpose in the argument.

Section B: Summary (8 marks)

Question 16 (8 marks)

Question: Using your own words as far as possible, summarise the risks and concerns associated with algorithmic governance, as outlined in the passage. Write as one continuous paragraph of no more than 120 words.

Content Points (for marking):

The following content points are drawn from lines 26–48:

1. Opacity/Black box problem: Algorithmic systems are difficult or impossible to understand, even for their creators, making it hard for affected individuals to challenge decisions.
2. Undermining justice: The inability to understand or contest algorithmic decisions violates the fundamental right to know the basis on which one is judged.
3. Bias amplification: Algorithms learn from historical data that contains past prejudices, and they replicate and potentially intensify these discriminatory patterns.
4. False neutrality: Algorithms appear mathematically objective but actually embed and perpetuate existing inequalities under a surface appearance of neutrality.
5. Accountability gap: When algorithms cause harm, responsibility is diffused among multiple parties (designers, deployers, data collectors), so no single actor can be held accountable.
6. Diffusion of responsibility: Each party involved can point to others, creating a situation where harmful outcomes occur without anyone bearing moral or legal responsibility.

Marking Scheme:

Language marks:

• 2 marks: Effective use of own words throughout; clear and concise expression.
• 1 mark: Some use of own words but with occasional lifting; generally clear.
• 0 marks: Heavy lifting from the passage; unclear or incoherent expression.

Sample Summary (for reference):

Algorithmic governance poses three major risks. First, many systems operate as black boxes, making their decisions impossible to understand or challenge, which undermines the right to justice. Second, because algorithms learn from historical data containing past prejudices, they replicate and intensify discrimination while appearing mathematically neutral — a false objectivity that masks embedded bias. Third, when algorithms cause harm, accountability becomes diffuse: designers, managers, and organisations each deflect blame, creating a gap where no one bears responsibility. Together, these risks mean that algorithmic systems, despite their efficiency, can perpetuate inequality and evade accountability unless properly regulated.

(Word count: 98)

Common Mistakes:

• Including the benefits/advantages of algorithms (not asked for — the question specifies risks and concerns only).
• Exceeding the 120-word limit.
• Lifting phrases directly from the passage without paraphrasing.
• Including the regulatory responses or the conclusion (outside the specified lines 26–48).

Section C: Application (7 marks)

Question 17 (7 marks)

(a) (4 marks) Identify and explain two concerns from the passage that are relevant to this scenario.

Answer:

Concern 1: Opacity / Black box problem The scenario describes a proprietary algorithm whose decision-making process is not disclosed to patients or doctors. This directly reflects the concern about opacity raised in the passage: when an algorithm operates as a black box, affected individuals cannot understand or challenge the basis of decisions. The 55-year-old patient's family is told the decision "cannot be overridden" and no explanation is given, which mirrors the passage's concern that opacity undermines the right to contest decisions.

Concern 2: Bias amplification The algorithm considers factors such as "age" and "past health behaviours" like smoking history. These factors could encode historical biases — for example, if past transplant decisions systematically disadvantaged older patients or those with certain health behaviours, the algorithm would replicate and potentially intensify this pattern. The scenario shows a 55-year-old being denied in favour of a 30-year-old with a "similar medical profile," suggesting that age may be functioning as a discriminatory factor, which aligns with the passage's warning about bias amplification.

Marking Scheme (per concern):

• 1 mark for correctly identifying a concern from the passage.
• 1 mark for explaining how it applies to the scenario.
• Total: 2 marks × 2 concerns = 4 marks.

Common Mistakes:

• Identifying concerns not found in the passage (e.g., "the algorithm might malfunction").
• Describing the scenario without linking it to specific passage concerns.
• Only naming the concern without explaining its relevance.

(b) (3 marks) Based on the ideas in the passage, suggest one measure that could address the concerns identified in part (a). Explain how this measure would help.

Answer:

Measure: Mandatory transparency / Right to explanation Following the model of the EU's GDPR mentioned in the passage, the hospital network could be required to provide patients with a meaningful explanation of how the algorithm reached its decision. This would address the opacity concern by making the decision-making process legible and contestable. If the 55-year-old patient's family could see the specific factors and weightings that led to the denial, they could identify potential biases (such as undue weighting of age) and challenge unfair outcomes. This measure would also help close the accountability gap, as transparency creates a record that can be audited and attributed to specific design choices.

Alternative acceptable measures:

• Independent audit/oversight body to review the algorithm's decisions for bias and fairness, addressing both bias amplification and accountability concerns.
• Removal or adjustment of problematic variables (e.g., reducing the weighting of age or health behaviours) to prevent bias amplification.
• Human override mechanism allowing doctors to appeal algorithmic decisions, addressing the accountability gap.

Marking Scheme:

• 1 mark for suggesting a measure grounded in the passage's ideas.
• 1 mark for explaining how the measure addresses the specific concern(s).
• 1 mark for demonstrating clear application to the scenario.
• Total: 3 marks.

Common Mistakes:

• Suggesting measures not supported by the passage (e.g., "ban all AI systems").
• Suggesting a measure without explaining how it would help.
• Being too vague (e.g., "make it fairer" without specifying how).

End of Answer Key

Total Marks: 50