AA map appears to offer certainty: coastlines, borders, roads and distances are reduced to marks on a flat surface. Yet the Earth is roughly spherical, so any attempt to represent it on paper or a screen requires distortion. Cartographers choose a projection — a mathematical method for transferring locations from a curved surface to a plane — and every projection preserves some properties while sacrificing others. The familiar problem is not that maps are inaccurate in a simple sense; it is that different kinds of accuracy cannot all be maintained at once.

BThe Mercator projection, developed in the sixteenth century, illustrates this trade-off. It preserves local angles, which made it extremely useful for sailors because a constant compass bearing could be drawn as a straight line. The cost is a dramatic inflation of area toward the poles. Greenland appears comparable in size to Africa on many Mercator maps, although Africa is roughly fourteen times larger. The projection was designed for navigation, not for comparing the territorial size of continents.

CModern mapmakers therefore match projections to tasks. Equal-area projections preserve relative surface size and are useful when the subject is population, forests or disease distribution. Other projections minimise distortion near a chosen country, which is valuable for national mapping. There is no universally 'correct' world map because the best choice depends on what users need to measure or compare. A projection can be technically appropriate for one purpose and misleading for another.

DDistortion is not limited to geometry. Maps also reflect decisions about selection. A transport map may enlarge a city centre and straighten railway lines so passengers can understand connections quickly. Harry Beck's famous London Underground diagram ignored geographic accuracy in favour of a simplified network of horizontal, vertical and diagonal lines. Stations became easier to follow, even though their apparent positions and distances no longer matched the streets above.

EDigital mapping introduced a different challenge. Online maps can change scale instantly and tailor what they display to a user's location. Businesses, landmarks and routes are ranked by algorithms, and two people may not necessarily see identical information. The map is no longer a single published object but a changing interface. This flexibility is useful, yet it makes the criteria for inclusion less visible: users may assume that what appears first is naturally the most important option.

FMaps can also influence political perception. The position of a border, the name assigned to a disputed territory, or the decision to centre a world map on one region rather than another may carry symbolic weight. Even when the underlying coordinates are identical, design choices affect which places seem central and which appear peripheral. Cartographers working for international organisations often use carefully negotiated labels because a seemingly minor naming choice can imply recognition of a political claim.

GPsychologists have found that repeated exposure to a particular map can shape mental geography. People often estimate places shown near the centre of familiar maps as more prominent or closer to them than they really are. Classroom maps may also reinforce assumptions about relative country size if students are not taught how projections work. For this reason, some educators deliberately compare several world projections rather than presenting one as neutral.

HNone of this makes maps untrustworthy. A useful map is a model, and models become powerful precisely because they simplify. Problems arise when the simplification is hidden or when a map designed for one purpose is interpreted as though it served another. Good map reading therefore requires a question that users rarely ask: not simply 'Is this map accurate?' but 'Accurate for what?'

AThe claim that groups can be remarkably intelligent is often traced to a country fair in 1906. The statistician Francis Galton examined hundreds of estimates of the weight of an ox. Individual guesses varied widely, but their average came surprisingly close to the true figure. Later researchers found similar effects in tasks ranging from estimating quantities to answering factual questions. The phenomenon became known as the 'wisdom of crowds': errors made in different directions can cancel one another when many independent judgments are combined.

BIndependence is crucial. Elena Marquez, a behavioural economist, argues that crowd accuracy depends less on the number of people than on the diversity of information they bring. If everyone relies on the same rumour, website or influential speaker, averaging many answers merely repeats one source. In her experiments, groups whose members first recorded private estimates generally outperformed groups that discussed the problem before committing to an answer.

CSocial influence can therefore reduce collective accuracy. In a well-known experimental design developed further by Tobias Kern, participants made numerical estimates and were then shown the estimates of others. After seeing the group, many revised their answers toward the apparent consensus. The range of responses narrowed, which made the group look more confident, but the average did not consistently become more accurate. Kern calls this the 'confidence trap': agreement increases while useful variation disappears.

DNot all interaction is harmful. Priya Nandakumar studies small teams solving complex problems for which no individual possesses all necessary knowledge. She finds that structured discussion can improve performance when members are assigned different evidence and must explain their reasoning. The benefit comes from information being pooled, not from compromise itself. Teams perform worse when high-status members speak first and others simply align with them.

EThe digital world has expanded the scale at which collective judgment can be gathered. Prediction markets, for example, allow participants to buy and sell contracts linked to future events. Prices can aggregate dispersed information rapidly. Researcher Samuel Reed notes that such markets sometimes outperform opinion polls, particularly when participants have incentives to update their views. But he warns that thin participation, manipulation or ambiguous questions can make market prices look more precise than the underlying knowledge warrants.

FAnother approach is to search for what cognitive scientist Mei Chen calls 'surprisingly popular' answers. Instead of selecting the option chosen by the largest number of people, the method asks participants both for their own answer and for their prediction of what others will say. An answer receives extra weight when it is more common than people expected. Chen's studies suggest that this can identify correct minority views when most participants share a plausible misconception.

GCollective intelligence is also shaped by how a problem is framed. If a question is emotionally charged or closely tied to identity, people may interpret new evidence defensively rather than independently. Under those conditions, adding more participants does not guarantee improvement. A very large crowd can reproduce the same systematic bias. Diversity of background is useful only if it produces genuinely different information or reasoning, rather than superficial demographic variety.

HThe practical lesson is not to replace experts with votes. Crowds are strongest when judgments are independent, information is dispersed and errors are uncorrelated. Experts remain essential when a task requires specialised knowledge, diagnosis or causal explanation. The most reliable systems often combine the two: expert-designed questions, independent individual judgments and transparent methods for aggregation. Collective intelligence is therefore less a mysterious property of large groups than a result of carefully designed conditions.