Reading Lab

IELTS Academic Reading Practice Pack 18

A premium Academic Reading set on district cooling, calibrated trust in AI, and the governance of deep-sea mining.

Question count

Time allowed

60 min

Passages

academicreadingfull mockurban infrastructurepsychologyocean governancetfngynngqa candidate

Exam panel

You have 60 minutes including answer transfer time. Submit once at the end or let the timer finish the exam automatically.

Time remaining

60:00

0 / 40 answers filled

Write only what the question requires. One extra word can still lose the mark.

After submission, you will see your raw score, estimated Academic Reading band, and the correct answers for every question.

Passage 1

District Cooling and the Hidden Logic of Urban Heat Management

Why district cooling can reduce peak electricity pressure in dense cities, and why its success depends on urban form, ownership, and timing rather than equipment alone.

A.A. Air conditioning is usually imagined as a room-by-room technology: an individual unit serving an apartment, office, or shop. District cooling works from a different premise. Instead of each building generating its own chilled air or water independently, a central plant produces cooling and sends it through a network to multiple buildings. The idea sounds merely technical, but in hot cities it has become a planning question. Advocates argue that where density, load patterns, and land values align, a network can cool urban districts more efficiently than thousands of isolated machines. Critics respond that the model locks cities into expensive infrastructure whose success depends on assumptions that may not hold for decades.

B.B. Much of the attraction lies in system efficiency rather than in any miracle of thermodynamics. Large plants can use equipment that performs better at scale, locate machinery where waste heat can be handled more intelligently, and shift some cooling production away from the moment each building experiences peak demand. In some systems, chilled water or thermal storage can be produced at times that are more favourable for the electricity grid and used later when the city heats up. This does not abolish energy demand, but it can alter its shape. The difference matters because peak stress on electricity networks often creates costs that are out of proportion to total annual consumption.

C.C. Yet infrastructure that looks efficient in engineering diagrams may falter in urban reality. District cooling requires pipe corridors, coordinated construction, long-term contractual arrangements, and a sufficient concentration of users. A sprawling suburb with dispersed buildings may gain little from a network that would be highly effective in a dense waterfront development, hospital cluster, or commercial district. Timing is equally decisive. It is easier to build shared energy systems into new developments than to excavate established streets and retrofit fragmented ownership patterns later. The best moment to plan such a network may be before a district exists in its final form.

D.D. Ownership and governance complicate the picture further. A building owner deciding between an individual chiller and a district contract will not necessarily prioritise the same things as a city planner concerned with grid resilience, land use, and long-term emissions. The private calculation may focus on price certainty, service reliability, and lease structure; the public calculation may include avoided peak loads or the opportunity to coordinate cooling with broader energy strategy. Networked infrastructure often creates this tension. System gains can exist without appearing compelling to each actor asked to join the system voluntarily.

E.E. There is also a risk of confusing centralisation with infallibility. A distributed landscape of individual units is inefficient in some ways, but it is also decentralised. A district cooling network, by contrast, creates new dependency on the operator, the distribution loop, and the long-term maintenance discipline of shared assets. If governance is weak, a network intended to improve resilience can concentrate vulnerability instead. This is not an argument against shared systems as such. It is a reminder that physical integration must be matched by institutional competence, transparent pricing, and credible service standards.

F.F. The strongest cases for district cooling are therefore highly contextual. Campuses, airports, business districts, and master-planned urban extensions often present the load density and coordination structure the model needs. In such settings, district cooling can reduce rooftop clutter, centralise maintenance, and support broader energy planning. But the model should not be treated as a universal urban virtue. Like many infrastructure ideas, it works best when matched carefully to spatial form, ownership structure, and the sequence of development. Its value lies less in technological glamour than in a disciplined fit between network design and city-making.

G.G. What district cooling ultimately reveals is that urban heat management is not only a matter of appliances. It is also a question of whether cities can plan energy demand collectively rather than leaving every building to solve the same problem alone. Where that coordination is possible, networks may offer meaningful advantages. Where it is not, the promise of efficiency can become a language of overbuilding. The challenge is to tell the difference early enough that the pipework follows a workable urban strategy rather than substituting for one. That judgement depends on load forecasts, land-use sequencing, governance capacity, and the political willingness to treat cooling as shared urban infrastructure rather than as a purely private consumer choice.

Matching Headings

Questions 1-5

Choose the correct heading for paragraphs B-F from the list of headings below.

Write the correct Roman numeral, i-viii, in boxes 1-5.

1. Paragraph B

i. Why a network changes the timing of demand rather than removing it
ii. A warning that decentralised systems are always superior
iii. Why retrofitting and density determine whether a network is viable
iv. A conflict between public system gains and private joining incentives
v. The claim that district cooling should replace all room-level equipment
vi. A reminder that centralisation can create concentrated weakness
vii. A narrowly defined set of conditions in which the model works well
viii. The argument that electricity demand no longer matters in hot cities

2. Paragraph C

i. Why a network changes the timing of demand rather than removing it
ii. A warning that decentralised systems are always superior
iii. Why retrofitting and density determine whether a network is viable
iv. A conflict between public system gains and private joining incentives
v. The claim that district cooling should replace all room-level equipment
vi. A reminder that centralisation can create concentrated weakness
vii. A narrowly defined set of conditions in which the model works well
viii. The argument that electricity demand no longer matters in hot cities

3. Paragraph D

i. Why a network changes the timing of demand rather than removing it
ii. A warning that decentralised systems are always superior
iii. Why retrofitting and density determine whether a network is viable
iv. A conflict between public system gains and private joining incentives
v. The claim that district cooling should replace all room-level equipment
vi. A reminder that centralisation can create concentrated weakness
vii. A narrowly defined set of conditions in which the model works well
viii. The argument that electricity demand no longer matters in hot cities

4. Paragraph E

i. Why a network changes the timing of demand rather than removing it
ii. A warning that decentralised systems are always superior
iii. Why retrofitting and density determine whether a network is viable
iv. A conflict between public system gains and private joining incentives
v. The claim that district cooling should replace all room-level equipment
vi. A reminder that centralisation can create concentrated weakness
vii. A narrowly defined set of conditions in which the model works well
viii. The argument that electricity demand no longer matters in hot cities

5. Paragraph F

i. Why a network changes the timing of demand rather than removing it
ii. A warning that decentralised systems are always superior
iii. Why retrofitting and density determine whether a network is viable
iv. A conflict between public system gains and private joining incentives
v. The claim that district cooling should replace all room-level equipment
vi. A reminder that centralisation can create concentrated weakness
vii. A narrowly defined set of conditions in which the model works well
viii. The argument that electricity demand no longer matters in hot cities

True/False/Not Given

Questions 6-9

Do the following statements agree with the information given in Reading Passage 1?

In boxes 6-9, write TRUE if the statement agrees with the information, FALSE if the statement contradicts the information, or NOT GIVEN if there is no information on this.

6. The passage says district cooling always reduces total electricity use in every kind of city.

7. The writer suggests new developments are often easier places to introduce shared cooling systems.

8. The passage states that building owners and city planners always value the same benefits.

9. The writer argues that network design can compensate for the absence of a broader urban strategy.

Sentence Completion

Questions 10-13

Complete the sentences below.

Choose ONE WORD ONLY from the passage for each answer.

10. District cooling may help by changing the ______ of electricity demand.

11. A dense hospital ______ is given as one setting where the model may work well.

12. A network can create dependency on the operator and the distribution ______.

13. The final paragraph warns against allowing pipework to substitute for an urban ______.

Passage 2

Calibrated Trust and the Human Use of AI Advice

Why the real problem in human interaction with AI advice is not trust alone, but whether trust is properly calibrated to the system's strengths and failure modes.

A.A. Public debate about artificial intelligence often treats trust as a binary condition. Either people trust a system too much and become complacent, or they trust it too little and fail to use a useful tool. Researchers in human factors have long argued for a more precise question. The issue is calibration. A person's confidence should rise or fall in proportion to what a system can actually do, under which conditions, and with which kinds of failure. Miscalibration can occur in both directions. Distrusting a reliable warning model wastes capability; over-trusting a brittle advisory system imports error into human decision-making with unusual speed.

B.B. This matters because many AI tools do not fail randomly. They perform unevenly across tasks, input quality, contexts, and user expectations. A system may classify routine cases accurately while becoming unstable when asked to handle rare events or ambiguous data. If users are shown overall accuracy alone, they may form a false mental model of consistency. That problem is amplified when the interface communicates confidence elegantly but opaquely. Smooth presentation can create the impression of underlying competence even where the model is extrapolating beyond what it has learned well.

C.C. One response is to increase transparency, but the word itself hides several possibilities. Users may need to know how a system was trained, what kind of data it expects, what evidence it is drawing on in one case, or what circumstances have historically produced error. Providing all of this at once may overwhelm rather than assist. The design challenge is therefore selective transparency: enough explanation to support judgement, not so much detail that the user mistakes documentation for comprehension. More information does not automatically produce better calibration.

D.D. Domain context changes everything. In navigation software, a small detour suggested by a model may be cheap to verify or ignore. In medicine, credit assessment, hiring, or judicial support, the cost of misplaced trust can be much higher and distributed unevenly across populations. This is why studies of AI reliance cannot simply ask whether users followed the advice. They must ask what the stakes were, what alternatives existed, and whether users understood the boundary between assistance and authority. The same behaviour can be prudent in one setting and negligent in another.

E.E. Organisations frequently complicate calibration by embedding AI into performance systems. When workers are evaluated partly on speed or compliance, they may lean on automated advice not because they genuinely trust it, but because resisting it carries procedural cost. A clinician who double-checks an alert, or an analyst who challenges a ranking, may appear slower than a colleague who accepts the output immediately. Under such incentives, recorded trust behaviour can mislead observers. Apparent confidence in the system may actually be confidence in the penalty for deviating from it.

F.F. Designers have experimented with confidence displays, uncertainty bands, selective prompts to verify, and interfaces that encourage users to compare outputs with external evidence. Some of these tools improve judgement, but only when users understand what the indicators mean. A confidence score can be read as a probability, a recommendation strength, or a polished graphic, depending on prior training and domain culture. Calibration is therefore partly educational. A system cannot rely on interface design alone if its users have not been taught how to interpret uncertainty in context.

G.G. The goal, then, is not maximal trust but appropriate reliance. Good human-AI systems help users know when to defer, when to inspect, and when to override. That requires more than technical accuracy. It requires interface choices, organisational incentives, training, and institutional humility about what the model does not know. A trustworthy system is not one that is always believed. It is one around which people can form dependable judgement. This is why the language of calibration matters so much: it shifts attention from whether a system feels impressive to whether people have learned to rely on it in proportion to its actual strengths, limits, and predictable failure modes in live settings. In practice, that means reliability is never a property of the model alone. It is a property of the relationship between the model, the task, the stakes, the interface, and the institution that teaches users what kind of disagreement with the machine is permitted.

Matching Information

Questions 14-17

Which paragraph contains the following information?

Write the correct letter, A-G, in boxes 14-17.

You may use any letter more than once.

14. the warning that a polished interface can exaggerate perceived competence

15. the idea that transparency must be selective rather than unlimited

16. the claim that observed reliance may reflect institutional pressure rather than genuine trust

17. the argument that calibration depends partly on user education

Matching Features

Questions 18-21

Look at the following elements and the list of statements below.

Match each statement with the correct element, A-D.

Write the correct letter, A-D, in boxes 18-21.

A. interface presentation

B. organisational incentives

C. domain stakes

D. user training

18. can make a model seem more stable than it really is

19. can push people toward compliance with automated advice

20. change whether following the same output is prudent or negligent

21. affects whether confidence indicators are interpreted properly

Multiple Choice

Questions 22-24

Choose the correct letter, A, B, C or D.

22. What is the main point of paragraph B? A. AI systems fail in entirely unpredictable ways. B. Users usually receive too much technical detail. C. Uneven performance can be hidden if people see only overall accuracy. D. Rare events are easier for models than routine cases.

23. According to the passage, why is more transparency not always helpful? A. Users prefer no explanation at all. B. Too much information can overwhelm judgement. C. Transparency only matters in navigation systems. D. Designers are unable to explain model training.

24. The writer's overall view is that strong human-AI performance requires A. maximum trust in systems with high average accuracy. B. removing all discretion from human users. C. appropriate reliance shaped by design, incentives, and training. D. banning confidence displays from interfaces.

Summary Completion

Questions 25-27

Complete the summary below.

Choose ONE WORD ONLY from the passage for each answer.

25. The passage argues that trust should be judged by how well it is ______ to system capability.

26. Users may form a false sense of model consistency if they see only overall ______.

27. A dependable human-AI system helps users decide when to defer, inspect, or ______.

Passage 3

Deep-Sea Mining and the Politics of Irreversible Evidence

Why debates about mining the deep ocean turn not only on metals and demand, but on what kind of evidence should be required before disturbance becomes effectively irreversible.

A.A. Deep-sea mining is often framed as a choice between two urgent transitions. On one side lies the argument that electric technologies require large volumes of critical minerals. On the other lies the claim that disturbing poorly understood ocean ecosystems in pursuit of those minerals would create damage that cannot be confidently bounded, repaired, or even fully observed. The dispute is not only environmental versus industrial. It is also epistemic. How much should governments claim to know before authorising extraction in a region where baseline ecological knowledge remains thin and direct recovery times may extend beyond institutional planning horizons?

B.B. Industry advocates tend to emphasise metal demand, technological progress, and the possibility of regulating extraction carefully. They note that nodules on the seabed contain materials already central to battery and industrial supply debates. Critics answer that the analogy to well-regulated land extraction is weak because the deep ocean remains more poorly mapped, more logistically inaccessible, and more difficult to monitor independently once operations begin. In this sense, the question is not simply whether harm can be reduced. It is whether governance institutions are being asked to guarantee a level of control that the evidence does not yet support.

C.C. Calls for a moratorium, precautionary pause, or outright ban are sometimes grouped together, but they are not identical positions. A ban treats extraction as unacceptable in principle. A moratorium usually links delay to conditions that might in theory be satisfied later. A precautionary pause may serve as a political signal that current knowledge, monitoring capacity, or rule-making remains inadequate. These differences matter because opponents of delay sometimes portray every restriction as absolute refusal, while supporters of delay may strategically blur distinctions that are important for law and diplomacy.

D.D. The institutional setting further complicates the debate. Decisions involve international seabed governance, contractor states, scientific advisers, private firms, and civil society groups with unequal power and information. A state sponsoring a contractor may have incentives that differ sharply from those of researchers emphasising uncertainty or from island governments worried about long-term marine impacts. Procedural fairness therefore becomes part of the environmental argument. Who gets to define sufficient knowledge, acceptable risk, and credible monitoring is not a neutral technical matter. It is distributed through institutions already shaped by strategic and commercial interests.

E.E. Some of the strongest pro-mining claims rely on comparison. If land-based mining also damages ecosystems and communities, the argument goes, then seabed extraction may be the lesser harm. That comparison may be politically potent, but it is not automatically decisive. Comparing two harmful supply options requires evidence not only about immediate disturbance, but also about traceability, labour governance, community consent, waste handling, and the possibility of substitution or demand reduction. A weak record on land does not by itself make a novel frontier acceptable. It may simply reveal that material policy has been poorly governed in more than one place.

F.F. For this reason, the debate increasingly centres on proof. Not proof that every impact is known in advance, which is unrealistic, but proof about thresholds: what range of disturbance is plausible, which effects are reversible, how monitoring would detect breach, and who would bear the burden of showing that a contractor remains within agreed limits. Where uncertainty is deep, the allocation of proof becomes a political choice. A governance regime can require critics to demonstrate likely catastrophe before delay is justified, or require proponents to demonstrate that monitoring and response capacities are credible before extraction proceeds.

G.G. Deep-sea mining thus compresses a wider problem of twenty-first century resource politics. States want minerals without bottlenecks, firms want new frontiers, and environmental institutions are asked to govern ecosystems whose most consequential features may still be poorly understood. The issue is not whether decision-makers can eliminate uncertainty. It is whether they are willing to treat uncertainty itself as a reason for restraint rather than as empty space to be filled later by commercial momentum. In that sense, the debate is a test of how institutions behave when commercial urgency arrives earlier than ecological knowledge and when law is expected to stabilise a frontier that science is still struggling to describe in adequate detail. The answer will shape not only this industry, but also the wider norm by which environmental governance decides when incomplete evidence is a prompt for caution and when it is treated merely as an obstacle to be overcome by administrative confidence.

Yes/No/Not Given

Questions 28-31

Do the following statements agree with the views of the writer in Reading Passage 3?

In boxes 28-31, write YES if the statement agrees with the views of the writer, NO if the statement contradicts the views of the writer, or NOT GIVEN if it is impossible to say what the writer thinks about this.

28. The writer thinks the dispute involves questions about knowledge as well as about industry and ecology.

29. The writer believes a moratorium and an outright ban are the same position in legal terms.

30. The writer says land-based mining has no relevance to seabed extraction debates.

31. The writer sees the burden of proof as politically important where uncertainty is deep.

Note Completion

Questions 32-33

Complete the notes below.

Choose ONE WORD ONLY from the passage for each answer.

32. The writer describes the disagreement partly as an ______ issue.

33. Critics argue that independent ______ after operations begin would be difficult.

Table Completion

Questions 34-35

Complete the table below.

Choose ONE WORD ONLY from the passage for each answer.

34. Position treating extraction as unacceptable in principle: ______

35. Procedural issue highlighted by the writer: unequal ______ and power

Flow-chart Completion

Questions 36-37

Complete the flow-chart below.

Choose ONE WORD ONLY from the passage for each answer.

36. Uncertainty remains deep -> governance must allocate the burden of ______ -> extraction may be delayed or allowed

37. If uncertainty is ignored, later commercial ______ may fill the gap

Diagram Labelling

Questions 38-39

Label the diagram below.

Choose ONE WORD ONLY from the passage for each answer.

38. Seabed mineral form mentioned in the passage: ______

39. One kind of government identified as worried about long-term marine impacts: ______ governments

Short-answer Questions

Question 40

Answer the question below.

Choose NO MORE THAN TWO WORDS from the passage for your answer.

40. What may environmental institutions be asked to govern despite poor understanding, according to the final paragraph?