IELTS Academic Reading Practice Test 445

On the desert coast of northern Peru, a few kilometres from the modern city of Trujillo, stand the remains of a city built almost entirely of mud. Chan Chan was the capital of the Chimú, a people who dominated a long stretch of the Pacific coast for several centuries before their conquest by the Inca in about 1470. It is the largest city of pre-Columbian South America and the largest adobe city ever constructed, spread across many square kilometres of desert. In 1986 it was added to UNESCO's World Heritage List.

The Chimú state emerged in the fertile river valleys of the north coast, in a landscape where rain almost never falls and where life depends on water carried down from the Andes. It grew from the collapse of the earlier Moche culture, whose ceremonial centres lie in the same valleys, and by around 900 its rulers had established themselves at Chan Chan. Over the following centuries the Chimú expanded northwards and southwards, absorbing neighbouring peoples, until they controlled a coastal strip of perhaps a thousand kilometres. Their wealth came from irrigated farmland, from the sea, and from the labour of the populations they subdued.

Farming on this scale required engineering. The Chimú dug canals that carried river water many kilometres to fields far from the valley floor, and they built sunken plots, known as huachaques, in which crops could be grown close to the water table. Maize, beans, squash and cotton were cultivated, while fishermen took to the ocean in small reed boats of a kind still used on the Peruvian coast today. Chimú metalworkers were among the most skilled in the Americas, casting and hammering gold, silver and copper into vessels, ornaments and tools.

The heart of the city is made up of nine or ten enormous rectangular compounds, usually called ciudadelas, or citadels. Each is enclosed by thick earthen walls that in places rise to nine metres, and each contains a maze of courtyards, corridors, storerooms and kitchens, together with a platform for burials and a well or reservoir. Archaeologists believe that each compound was built as the palace of a ruling king during his lifetime and continued to be used, after his death, as a store for tribute and as a base for the administration of his estates. Beyond the compounds, ordinary houses of cane and mud spread across the surrounding desert.

Chan Chan is best known for its decoration. The walls of the compounds were once covered with friezes modelled in damp clay, and although centuries of wind and rain have destroyed much of this work, long sections survive. The designs are abstract and repetitive: waves, nets, fish, seabirds and small geometric figures arranged in bands. Because the same motifs recur across the site, some scholars argue that they carried meanings connected with the sea and with fertility rather than serving purely as ornament.

Building on such a scale raised practical problems. The Chimú made their walls from adobe, bricks of clay, sand and water shaped in wooden moulds and dried in the sun, laid in thick courses and bonded with mud mortar. Where a wall had to be especially massive, the builders sometimes packed damp earth between boards and rammed it hard. The work was done in stages: each section was added, allowed to dry and then decorated, which is one reason why the ruins show so many joins and repairs. Nothing was fired; the whole city depended on the absence of rain.

The end came from the mountains. Around 1470 the Inca, whose empire was expanding rapidly under Tupac Inca Yupanqui, moved against the Chimú. According to accounts written down after the Spanish conquest, the Inca cut the canals that supplied the city with water and forced its ruler, Minchançaman, to surrender. The defeated king was taken to Cusco, and Chimú goldsmiths and other craftsmen were sent there too, because their skills were valued by the Inca court. Chan Chan survived as a provincial centre, but its palaces were no longer maintained.

After the Spanish arrived in the 1530s the city's decline quickened. The town of Trujillo was founded nearby in 1534, and for generations its inhabitants helped themselves to the ready-made bricks of Chan Chan, while treasure hunters searched the burial platforms for gold. By the time scholars began to study the ruins seriously in the nineteenth century, much of the site had already been quarried away.

Today the greatest danger is water. Chan Chan was built in one of the driest places on earth, but the modern irrigation schemes that water the fields around it have raised the water table, and salty groundwater now rises into the bases of the walls, where crystals form and crumble the clay. Occasionally the climate itself turns against the city: the warm ocean current known as El Niño brings heavy rain to this coast every few years, and the storms of 1998 and 2017 stripped the surface from walls and friezes.

Conservation is a slow business. Teams of archaeologists and conservators work under temporary roofs, recording and stabilising one section at a time, and visitors follow a fixed route that keeps them away from the most fragile areas. In 1986, the year of its inscription, the site was also placed on the list of World Heritage in Danger, on which it remains. Protecting a city made of dried mud has proved to be a long campaign.

Most people would say that they choose what to have for breakfast, whether to walk or drive to the shops, and when to check their phones. In reality a large share of such behaviour is not chosen at all. Psychologists define a habit as an action that has become automatic: it is triggered by something in the surroundings, such as a time of day, a place or the end of another activity, and it is carried out with little conscious thought. One widely cited estimate, based on a study in which people recorded their everyday activity, is that around forty per cent of what we do each day is repeated in the same context.

The idea that habits have a structure was popularised in the 2010s by writers who, drawing on earlier research, described a loop of cue, routine and reward. A cue sets the behaviour going; the routine is the behaviour itself; the reward is what makes the brain repeat it another time. At the level of the nervous system, this repetition is thought to depend on the basal ganglia, a group of structures deep in the brain that are involved in learning sequences of movements. Studies in animals suggest that the link between a cue and an action may be stored somewhat separately from the action itself.

How long it takes to form a habit is often said to be three weeks. The figure appears to come from a plastic surgeon, Maxwell Maltz, who noticed in the 1950s that patients needed at least about twenty-one days to grow used to a changed appearance; the number was later borrowed for behaviour of every kind. A far more careful answer came from a study at University College London published in 2009. Ninety-six volunteers were asked to adopt a new eating, drinking or activity habit in a fixed situation and to record how automatic it felt each day. The average time taken to reach the point where the behaviour felt automatic was sixty-six days, but individual times ranged from eighteen days to more than eight months.

If habits are triggered by context, then a change of context should disturb them. This has been shown in several ways. In one well-known experiment, people who regularly ate popcorn at the cinema were given stale popcorn in a cinema; those who were strongly in the habit ate just as much of it as they would have eaten of fresh popcorn. Given the same stale popcorn in a meeting room, they ate much less. Students moving to a new university show a similar pattern: their old habits weaken, and new ones, sometimes more desirable ones, can be established more easily during the first weeks after the move.

Most researchers regard repetition and reward as the two main ingredients of a habit, although some argue that repetition alone can be enough. A behaviour repeated without any satisfying outcome tends to fade, and a single pleasant experience does not create a habit. Laboratory work with animals has shown that the brain responds to unexpected rewards, and that these responses help to strengthen the connection between a cue and an action. The effect builds up gradually, which is why habits form at different speeds and why an occasional lapse does not, in itself, undo the learning.

Breaking an unwanted habit is harder than making a new one, largely because the cue is still there. Suppressing the behaviour by force of will is rarely successful in the long run; it is tiring, and the effort tends to collapse precisely when people are stressed or distracted. Two approaches have better support. The first is to alter the environment so that the cue disappears — keeping snacks out of the house, for example, rather than resolving not to eat them. The second is to plan a replacement in advance, in the form of an if-then rule: if I finish lunch, I will go for a short walk. Experiments on such plans, known as implementation intentions, suggest that they work better than general resolutions.

Much remains uncertain. Most studies of habit rely on people's reports of how automatic their behaviour feels, and these reports may be shaped by what participants expect to experience. The contexts that trigger habits are difficult to measure outside the laboratory: a cue such as finishing a meal is not a simple stimulus, and two people may respond to the same situation quite differently. It is also unclear how far the results of any one study can be generalised, since volunteers are often students and the habits studied are usually trivial ones. Even the idea of an average formation time, useful as it is, conceals a very wide spread of individual experience.

Practical advice follows from this research, though it should be treated with caution. Because context matters, the most effective moment to change a habit is often a change of scene: a new home, a new job or a holiday. Because repetition matters, small and frequent practice is more useful than occasional intense effort. And because reward matters, a new behaviour is more likely to stick if it brings some immediate pleasure or relief, rather than a benefit that appears only months later. None of this is a formula, but it is a more realistic picture than the promise of a habit in twenty-one days.

A large graduate employer may receive several thousand applications for a few hundred places. No team of humans can read them all with equal attention, so a growing number of organisations use software to sort, score or reject candidates before a person ever sees them. The programs range from simple checks for particular qualifications to machine-learning models that rank applicants on hundreds of variables, and to video systems that claim to assess personality from the way someone speaks. Whether this makes hiring fairer or less fair is now a matter of genuine dispute, and the arguments on both sides deserve to be taken seriously.

Supporters of algorithmic screening begin with the weaknesses of the alternative. A single interviewer, meeting a candidate for forty minutes, forms an impression quickly and then looks for evidence that confirms it; two interviewers may reach quite different conclusions about the same person. Because the questions asked and the standards applied vary from one conversation to the next, the traditional interview is a poor predictor of how someone will do the job. A standardised assessment, by contrast, applies the same rule to every applicant and can be checked afterwards. Its advocates also point out that a model can be audited, whereas an interviewer's reasons for rejecting a candidate usually go unrecorded.

Critics make two objections. The first concerns the data on which such models are trained. A system learns what a good employee looks like by studying the company's existing staff and its past decisions, so any bias in those decisions is likely to be reproduced. One large technology company is reported to have abandoned an experimental recruitment tool after it began downgrading CVs that mentioned women's activities, apparently because it had been trained on a decade of applications from men. The model was not malicious; it had simply learned the pattern it was shown.

The second objection concerns the individual. A candidate rejected by an automated system is rarely told why, and has no practical way of challenging the decision. Even the employer may be unable to explain it, since the relationships a machine-learning model has found in the data cannot easily be translated into reasons. A rejection that cannot be explained is difficult to appeal against, and the use of such tools therefore raises a question of fairness that has nothing to do with statistics.

Regulators have begun to respond. In Europe, the Artificial Intelligence Act treats software used to select or rank job candidates as high-risk, which means that employers must document how it works, keep records, and ensure that a human being remains in control of decisions. New York City has taken a narrower route: since 2023, employers there have had to have such tools audited for bias and to tell candidates when one is being used. Both approaches accept that the technology will be used, and try to make its use accountable.

Meanwhile, researchers who study how people respond to algorithms have found a pattern that complicates the debate. In experiments, participants who see an algorithm make an error lose confidence in it far more readily than they lose confidence in a human who makes the same error, even when the algorithm has proved the more accurate of the two. Hiring is unusual, however, because the people who make the decisions rarely receive clear feedback about whether their choices were right, so confidence in human judgement is seldom corrected by experience.

Firms that sell recruitment software defend their products vigorously. They argue that their assessments measure qualities — problem-solving, attention to detail, resilience — that are relevant to the job and that candidates cannot easily fake, and that the alternative is a manager's intuition. Some also note that clients demand speed, and that a system which removes obviously unsuitable applicants saves time that can be spent on the rest. What is largely missing from these claims is independent evidence: commercial confidentiality is often cited as a reason for refusing outside scrutiny.

Independent evidence is scarce for a further reason. Employers seldom publish the outcomes of their selection systems. The connection between being hired and performing well becomes visible only years later, by which time the tool has usually been replaced. Candidates, for their part, quickly learn what a test rewards, so a measurement that once captured ability may end up capturing coaching. None of this means that the tools do not work; it means that nobody outside the companies selling them is in a position to say.

My own view is that the choice between a human and a machine is the wrong way to frame the question. Automated screening can be justified when it is limited to structured, job-relevant criteria, when applicants at the borderline are reviewed by a person, and when the employer can show that the tool predicts performance in the job rather than merely resembling the staff it already has. Where those conditions are not met, a tool that cannot be explained or challenged should not be used. The human procedures now being replaced are themselves far from reliable, and the strongest argument for automation may be that it forces companies to say, at last, what they are looking for.

On that view, the answer to the question in the title is a qualified one. Employers may use artificial intelligence to choose among applicants, but the burden of proof should rest with them rather than with the people who are turned down.