eduKate Singapore

VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

Next useful routeCanonical hubHELP

How Intelligence Works | Identifiable Victim Effect — Stories, Statistics and Resource Decisions

—

by

eduKate Asia
Three students in school uniforms work through open books at a classroom table, with textbooks and stationery nearby and study notes on the whiteboard behind them.

The identifiable victim effect describes a possible difference in how people respond to a named or vividly presented person in need versus an anonymous description of people at risk. A single face or story can make the need concrete while a larger statistical total may remain emotionally distant. The practical lesson is not to distrust compassion. It is to use stories for attention and dignity while using comparable evidence, fair criteria and explicit denominators to decide how limited help should be allocated.

The evidence is more complicated than a slogan that one named person always matters more than thousands unnamed. Influential experiments reported identifiable-victim effects, but a preregistered replication and reanalysis published in Collabra: Psychology found no support for the effect in its tested hypothetical-donation conditions and raised the possibility that insensitivity to the number helped is a better explanation in some settings. Intelligent decisions should therefore treat identifiability as a plausible framing influence to test, not a universal law of the human mind.

Contents

Why a face can change the question

Imagine a community programme deciding how to spend a fixed support budget. One appeal introduces an identified child and describes a specific unmet need. Another describes a group of families and gives only counts, percentages and average outcomes. The first appeal may make it easier to picture an individual; the second may better show the scale of the underlying problem. A thoughtful decision needs information from both.

The cognitive distinction is between representation and allocation. A story represents experience in a format human attention readily understands: a protagonist, obstacle, time, setting and possible action. Statistics represent a distribution of cases that no one can view simultaneously. They answer a different question: how many people are affected and what a particular intervention could achieve across them. Problems arise when the emotionally vivid representation is permitted to stand in for comparative evidence.

The point is not that a named child is somehow less real than a large count. Each person in the count is equally real. The difficulty is that the mind cannot imagine each member of a population with the same vividness as one described individual. However, vividness is also affected by culture, proximity, language and prior experience. Do not assume that every audience responds in the same way.

A simple repair is to separate two stages. Let the story clarify what harm feels like and why action matters. Then compare interventions using a common outcome measure, feasible cost, uncertainty and fairness constraints. A powerful narrative can motivate attention without being allowed to overwrite the arithmetic.

What the research actually proposes

Early research on identifiable and statistical victims explored whether donors respond differently when a person is identified and when the same general need is represented statistically. Small, Loewenstein and Slovic examined donation choices and the role of deliberative thought. Related work discussed psychic numbing: the possibility that increasing the number of people at risk does not generate a proportionate emotional response.

These are claims about decision tendencies in particular settings, not a statement that people invariably choose a named person over a group. Research designs differ in whether donations are hypothetical or real, whether the victim’s identity is a name, photograph or personal description, and whether the control condition involves one anonymous person or many. Changing those details changes the question being tested.

The 2023 Collabra: Psychology replication, shown online with an article date in January 2024, reported no empirical support for the identifiable-victim effect in its tested conditions using hypothetical donations. Its authors also raised publication-bias concerns in the earlier literature and suggested scope-insensitivity as a potentially useful interpretation. They explicitly cautioned about what their findings do and do not establish.

A reader should therefore avoid treating older and newer results as a simple vote count. The responsible question is which construct was measured, how the cases were framed, whether money was real or hypothetical, how large the sample was and how much uncertainty remained. The general educational value is stable even while effect sizes are debated: attention and evidence are different inputs to a resource decision.

Attention is not the same as ethical importance

An identifiable person can attract attention more readily because the description is concrete. Yet ethical importance is not determined by how memorable a profile happens to be. People without compelling photographs, fluent advocates or public visibility can have needs just as severe. A decision procedure becomes inequitable if it systematically favours cases that happen to communicate better.

Consider two equally urgent cases. One has a detailed personal narrative; the other is recorded only in a short administrative entry because the family lacks an advocate. If the decision-maker assigns all funding to the first case on the basis of vividness, the selection mechanism has become a communication contest. Good intentions do not remove that procedural problem.

At the same time, attempting to eliminate every story can make a programme emotionally inaccessible. A purely statistical report may fail to communicate what an abstract outcome means in lived experience. The goal is not to suppress individual testimony but to prevent it from serving as a hidden substitute for eligibility and expected benefit.

The operational solution is transparent role separation. First-person evidence can identify overlooked needs, improve service design and test whether official categories describe real experiences accurately. Allocation rules should then specify who qualifies, what outcomes matter, how uncertainty is handled and what appeals or corrections are possible. Compassion helps define the problem; consistent reasoning helps distribute the response.

The difference between identifiability and scope insensitivity

Identifiability concerns whether a specific person’s representation changes the response. Scope insensitivity concerns whether the response changes appropriately when the number of people affected changes. Both can produce an uncomfortable pattern—strong concern for one case alongside weak sensitivity to a much larger problem—but they are not the same mechanism.

Suppose an appeal describing one anonymous learner receives an intended donation of 20 currency units, and a second appeal describing ten anonymous learners also receives 20 units. That pattern would indicate insensitivity to the tenfold difference in scope under those conditions. It would not prove an identifiable-victim effect because neither appeal named a particular person. A third condition identifying one learner would be needed to test identifiability.

Now suppose a named learner receives 30 units while one unnamed learner receives 20 units. That difference may be compatible with an identifiability effect, but the design must rule out other changing features. Perhaps the named case included more severe need, a clearer intervention or more evidence of efficacy. Fair comparisons hold those dimensions as constant as possible.

The distinction matters for repair. If the issue is identifiability, balancing narrative presentation across alternatives may help. If the issue is scope insensitivity, a structured comparison of how benefits scale with the number reached may be more useful. A single instruction to “think more rationally” is too vague to diagnose either problem.

A hand-checked funding example

Take an entirely fictional programme with a fixed budget of 1,200 currency units. Intervention A supports one learner at a total cost of 600 units and is expected to deliver an additional 24 tutoring hours to that learner. Intervention B costs 1,200 units and is expected to deliver 10 additional tutoring hours to each of four learners. These are illustrative planning assumptions, not measured effects of actual tuition.

A’s cost per additional tutoring hour is 600/24 = 25 currency units per hour. B’s planned total is 4×10 = 40 additional learner-hours, so its cost per additional learner-hour is 1,200/40 = 30 units per hour. On that narrow metric, A is cheaper per learner-hour. But A reaches one learner, whereas B reaches four. There is no mathematically valid answer until the programme states whether it values learner-hours, coverage, severity of need or another outcome.

The budget also changes the comparison. With 1,200 units, the programme could fund two instances of A if there are two eligible learners and the offer is repeatable, yielding 48 learner-hours at 1,200 units. That would be 1,200/48 = 25 units per learner-hour and cover two learners. If only one learner qualifies for A, that repeatability assumption fails and the unused funds require another use.

An identifiable appeal might make A feel compelling because the learner’s obstacle is vivid. A fair allocation review should explicitly list the feasible alternatives and denominators instead of inferring an answer from the story. The arithmetic exposes the trade-off; an ethically defensible decision still requires a value judgement about the outcomes.

Why denominators are essential

Statements such as “most of our help goes to this case” or “the project improves learning” are incomplete without specifying denominators. Is the denominator all applicants, all eligible learners, all funded learners, all learners completing the programme, or all hours delivered? Each supports a different interpretation and can change over time.

Suppose a pilot reports that eight of ten participants improved on a chosen test, an 80 percent observed improvement share. That fraction says nothing about twenty additional eligible learners who were never selected. If the original applicant pool contained thirty learners, eight improvements out of thirty applicants is about 26.7 percent. Neither number directly estimates causal impact without a valid comparison or experimental design.

Identifiable stories can hide denominators by focusing on the individual who benefited. But an aggregate table can also hide exclusion by counting only completed cases. Intelligent evaluation protects both the numerator and the opportunity set. It asks who was missing, why they were missing and whether the intervention could have benefited them.

When reporting a case study, state precisely that it is a case study. Pair it with programme-level outcomes that use stable definitions and disclose missing data. If a learner’s progress is genuine, it need not be exaggerated into evidence that all learners will achieve the same result. Transparent denominators honour the subject of the story more than an inflated claim would.

The role of emotion in reasoning

Emotion is not an enemy of intelligence. It can signal urgency, help a person notice potential suffering and motivate action that would otherwise be postponed. In many contexts, a statistic becomes meaningful only when connected to an understandable life situation. The challenge is that affective intensity and expected benefit are different dimensions.

Imagine two urgent medical services. One story makes a problem instantly imaginable. Another concerns preventive care that reduces risk across many unseen people. If a donor judges solely by what produces the strongest immediate emotional response, prevention may be undervalued. Yet if an analyst ignores human experience entirely, the programme may define success through convenient metrics that miss dignity, pain or autonomy.

A mature decision process allows emotional information to influence which outcomes matter while keeping comparative claims open to verification. Emotions may help answer “Why should we care?” They do not alone settle “Which feasible action helps most, under which constraints?” Both questions are legitimate and neither should pretend to replace the other.

The appropriate response is structured dialogue between experience and measurement. Listen carefully to affected people; assess evidence quality; make the value choices visible; and test whether the selected allocation would still look justified if the beneficiaries’ names and images were removed. That last test does not solve ethics, but it detects one way presentation can overwhelm consistent judgement.

Stories are data of a different kind

A case narrative can reveal a mechanism that aggregate outcomes miss. It might show that a learner could not attend sessions because of transport cost, that a parent misunderstood a notice, or that a public service required a form unavailable in the family’s language. Those observations can generate hypotheses about barriers, interventions and measurement gaps.

However, the frequency of a barrier cannot be inferred reliably from one compelling case. The story may be typical, unusual or deliberately selected. Before redesigning a whole programme, gather evidence about how common the barrier is and whether the proposed fix transfers across contexts. A qualitative account can guide what to measure; a suitable quantitative design can estimate how widely the pattern occurs.

Treat individual testimony respectfully without treating it as a substitute for population evidence. People should not be forced to disclose private trauma to qualify for services where other evidence would suffice. An individual’s dignity is not proportional to how emotionally affecting a story they can produce.

A practical evidence record distinguishes what the person said, what staff observed, which interpretation was inferred and what broader frequency remains unknown. This separation prevents the story from being dismissed as “anecdotal” while also preventing it from becoming an unsupported universal explanation. The best institutions can hear a person’s account and still maintain reliable allocation criteria.

Selection bias enters before the donation

Who becomes identifiable? Often someone who has access to a camera, a storyteller, a public platform, a spokesperson or an organisation skilled at promotion. This can select for communication resources rather than severity of need. If attention drives funding, groups already better connected to visibility may accumulate further advantage.

Selection also happens when organisations choose success stories after a programme ends. A learner who improved dramatically may be featured, while equally eligible learners who did not improve remain unnamed. Readers then encounter an apparent causal story built from the most favourable outcome. That is survivorship-style selection, not proof that the intervention caused the reported success.

A responsible campaign can still present a real example, provided it clearly labels the case, secures appropriate consent, protects sensitive details and pairs the narrative with honest programme-level evidence. Showing one person is not deceptive by itself. The misleading move is to imply that an unrepresentative case describes everyone’s likely outcome.

For the evaluator, the test is straightforward: would the same resource decision be reached if all cases had equal publicity? If not, identify whether the difference is justified by relevant evidence or merely by presentation. Asking this question before allocations can reduce both media-driven distortion and institutional blindness toward quiet needs.

Make the comparison set visible

A decision can be poor not because it misjudges the visible case but because it never considers viable alternatives. A single identifiable case arrives as a complete story; alternatives may remain abstract, unnamed or absent from the meeting. The choice then becomes “help this person or do nothing,” even though the real choice may involve many possible allocations.

Build a comparison set before deciding. Specify the available budget, the eligible population, the intervention options and the evidence for their likely effects. State the constraints that cannot be traded away, such as legal rights, urgent safeguarding needs, consent requirements or equitable access. Then compare options across the chosen criteria rather than constructing criteria after the most vivid case has won.

For example, a school welfare committee may have to choose among transport support, equipment loans and short-term tuition assistance. These options relieve different barriers. A persuasive narrative about equipment does not establish that equipment is the binding constraint for the whole applicant group. Conversely, an average outcome from a transport programme does not prove it is suitable for a learner whose difficulty is unrelated to travel.

The comparison set is not just a spreadsheet. It is a safeguard against framing. It gives the decision-maker a legitimate way to care about the identified person while recognising other people who may not have entered the conversation. The selection process becomes inspectable rather than dependent on who happened to speak first.

Quantify benefits without pretending everything is commensurable

It is tempting to reduce every outcome to a single score. That can be useful when the programme’s goals are well defined, but moral and educational outcomes are often multidimensional. Additional learning time, reduced stress, improved safety and family autonomy are not automatically interchangeable. A model can rank options, but the weights attached to different outcomes represent human choices.

Start with a small set of observable outcome measures and keep their units clear. An intervention might improve school attendance by an estimated number of days, reduce travel time by minutes per week, or increase the proportion of learners able to access a required service. These metrics inform a decision but do not prove that the most efficient intervention on one measure is ethically best.

Where evidence is uncertain, present ranges rather than false precision. If an intervention is estimated to help between five and nine learners, report the range and the assumptions rather than inventing a single guaranteed count. A vivid story can tempt decision-makers to substitute certainty of feeling for certainty of outcome; disciplined uncertainty reporting counters that temptation.

Finally, make explicit when a criterion is a floor rather than a weighted trade-off. For example, urgent protection from harm may be a priority obligation rather than a competing option in an ordinary cost-per-outcome ranking. Intelligence does not remove ethical disagreement; it makes the factual and normative components of that disagreement easier to see.

An original sensitivity analysis

Take two fictional interventions. A is estimated to deliver 30 units of relevant benefit at a cost of 600 currency units. B is estimated to deliver 48 units of benefit at a cost of 1,200 units. Their simple benefits per currency unit are 30/600 = 0.05 and 48/1,200 = 0.04 benefit units per currency unit. A appears more efficient under this invented common metric.

But suppose the estimate for A is highly uncertain: its plausible benefit range is 12 to 36 units. B’s plausible range is 42 to 54. A’s efficiency range is 12/600 = 0.02 to 36/600 = 0.06, while B’s is 42/1,200 = 0.035 to 54/1,200 = 0.045. The ranges overlap, so a simple point-estimate ranking hides uncertainty.

Now change the decision target from efficiency to reaching at least 40 benefit units with one funded intervention. Under the assumed intervals, A cannot meet 40; B may. The recommended option changes because the objective changed, not because a story became more persuasive. If the budget is only 600 units, B is infeasible without additional funding, and the choice set changes again.

The calculation illustrates a general intelligence skill: separate what the evidence says from what the decision is trying to maximize or guarantee. Identifiability may influence which case receives attention, but robust decisions need explicit constraints, sensitivity to assumptions and honest uncertainty about the underlying benefit model.

A fair way to use a personal appeal

An organisation can present a case narrative without implying that the named person is more deserving than everyone else. Begin with the person’s actual experience, using informed consent and respectful language. Then explain the systemic problem that the case illustrates, how many comparable people may be affected and what is known or unknown about intervention outcomes.

Avoid unsupported equivalences such as “your contribution will save one life” when the evidence only supports a partial or probabilistic benefit. Likewise, do not use photographs or emotionally loaded detail that the person did not consent to share. Identifiability should improve understanding and connection, not turn someone’s difficulty into compulsory public exposure.

Give the reader a route from feeling to action. If donations go into a common fund, say so. If the contribution is earmarked for one individual, identify the restriction and what happens if that need is already met. If the programme uses triage rules, explain them. Transparency lets supporters care deeply without assuming that the most visible case necessarily receives every marginal unit.

Finally, evaluate campaign effectiveness with suitable outcome measures. More clicks, tears or shares are not the same as better resource allocation. Did the campaign secure useful support? Did the allocation reach eligible people? Were claims accurate? Did any group become less visible or less protected? These measures connect the emotional doorway to the actual consequences of giving.

How teachers can teach the distinction

A classroom can use an age-appropriate thought experiment without exposing students to distressing real-world imagery. Present a fictional case about one learner needing study materials, followed by a second scenario involving multiple learners with comparable needs. Ask students first what each scenario makes them notice; then provide costs, eligibility, possible outcomes and a fixed budget.

The first task is descriptive: what information is vivid, missing or emotionally salient? The second is mathematical: calculate feasible combinations and benefits using stated units. The third is ethical: explain which criteria should matter and which should not. The fourth is reflective: would the group make the same decision if identities and presentation order changed?

This exercise teaches that arithmetic alone cannot settle moral priority, while emotion alone cannot reveal all trade-offs. Learners practise holding both kinds of information without conflating them. A student may reasonably prioritise urgent need over expected aggregate benefit, provided the value choice is stated clearly and applied consistently.

Teachers should assess reasoning rather than reward one predetermined ethical answer. A strong response identifies constraints, computes quantities correctly, recognises unseen stakeholders and acknowledges uncertainty. If students revise their preferences after seeing a fair comparison, that is a demonstration of learning, not a sign their initial compassion was foolish.

Group decisions can amplify a single story

Teams do not automatically correct individual framing effects. One member may tell a vivid story early in a meeting, creating a shared reference point. Other evidence may be introduced later as a table that receives less discussion. Social agreement can build around the first narrative before the group has asked what alternatives exist.

A practical countermeasure is independent pre-meeting judgement. Ask each participant to evaluate anonymous case summaries or common outcome criteria before hearing the full presentation. Then compare decisions, show the resource constraints and invite specific disconfirming evidence. This reduces the risk that an influential speaker or presentation order becomes an unnoticed allocation rule.

The team should also distinguish the emotional value of hearing a case from the evidential value of selecting it. Testimony can reveal facts that a summary omitted, and those facts may legitimately alter the decision. But the same criteria should then be used for comparable unpresented cases where evidence can be gathered. The correction is not to exclude the story; it is to prevent selective standards.

Institutional design matters most where the stakes are recurring. A transparent appeals process, written eligibility definitions and review of unfunded cases can reduce persistent visibility bias. A well-structured meeting helps today; a stable rule improves decisions over years. Collective intelligence is measured by whether relevant information changes the final allocation, not merely by whether participants felt moved.

Distinguish the effect from neighbouring biases

The identifiable victim effect overlaps with several other psychological ideas, but its precise claim should not dissolve into a general warning against emotion. Availability refers to judging frequency or likelihood through easily retrieved examples. Survivorship bias arises when attention focuses on cases that remain visible after a selection process. Denominator neglect occurs when a compelling numerator is considered without its relevant comparison total.

Outcome bias is different again: judging the quality of a prior choice largely from how it turned out rather than from the evidence available beforehand. A vivid success story might trigger all of these processes, but an analyst should identify which one the decision actually displays. Calling every memorable appeal an identifiable-victim effect makes the diagnosis vague.

For instance, if a committee funds one named person despite a stronger anonymous alternative, identifiability is a candidate mechanism. If it overestimates how common a dramatic problem is after seeing one example, availability may be more relevant. If it features only programme winners, selection bias may dominate. If it reports seven helped without identifying the twenty eligible, denominator neglect is central.

Different mechanisms require different repairs. Match narratives across conditions to test identifiability; compare base rates to address availability; sample the full applicant pool to address selection; state all relevant denominators to address misleading rates. Precise vocabulary supports precise institutional correction.

The replication finding changes the responsible claim

A 2024 online article presenting a 2023 Collabra: Psychology study revisited the influential Small and colleagues experiments using a preregistered replication and extension. The paper described a reanalysis of earlier meta-analytic evidence raising publication-bias concerns. In its tested hypothetical-donation scenarios, it reported no support for the classic identifiable-victim effect.

That does not logically establish that identifiable appeals never influence real donations, different populations or other decision contexts. Nor does it mean every original finding was necessarily false. Replication evidence narrows confidence in a sweeping claim and encourages investigators to distinguish identifiability from other constructs, including scope insensitivity and perceived impact. Methodological differences can matter.

A good educational article must reflect that state of evidence. It can describe the proposed mechanism and use it as a diagnostic hypothesis while avoiding categorical language such as “people always donate more to one named victim.” Readers deserve to know that a famous psychological label may have a smaller, more context-dependent or less reproducible effect than popular summaries suggest.

The practical safeguards remain valuable independently of a precise effect size. Transparent denominators, common eligibility criteria, consent-based storytelling and comparison of available interventions can improve allocation decisions even if identifiability makes little difference in a particular population. That is an important feature of sound reasoning: the recommended procedure should not depend on exaggerating a contested research result.

Changing context: prioritising a school support queue

Suppose a school support team receives a vivid request from one family for urgent equipment, while five other cases arrive as short anonymous records. The team has only enough capacity to resolve two cases this week. The emotionally salient route is to choose the detailed case immediately. A procedural route begins by defining urgency, harm if delayed, feasibility, prior waiting time and any safeguarding obligations.

The team does not need to remove personal testimony from the process. It can ask the same clarifying questions of each case and record the answers in a comparable format. A learner’s narrative may reveal an urgent accessibility need that was absent from the short administrative entry; if so, that is relevant new evidence. Anonymisation of the comparison is useful only after the material facts have been captured fairly.

Now change the setting. Instead of six equipment requests, the school must decide between upgrading a shared accessibility facility and providing one individual device. The individuals affected, time horizon and feasible outcomes differ. Comparing a compelling one-person story with an abstract future benefit is harder, but the same structure applies: list stakeholders, define outcomes, state uncertainty, calculate resource constraints and make value choices explicit.

The explained answer is not “always pick the group” or “always pick the person.” It is to ensure that the identity and storytelling quality of an applicant do not silently become the decision criterion. People may legitimately need different support; the criteria for recognising that need should be reasoned and consistent.

AI tools can reproduce the presentation problem

A summarising AI system can unintentionally reinforce visibility bias. If one case contains a long human story and ten other records contain short structured fields, a generated executive summary may allocate most of its space to the narrative. A manager reading only the summary can then overestimate the exceptional case’s relative importance.

A better workflow instructs the system to preserve the full denominator and to distinguish selected examples from the population. The summary can introduce one consented, clearly labelled example, then provide a comparable table of all cases and a separate discussion of missing data. Important allocation recommendations should be checked against source records and a human-owned policy.

Retrieval systems face related selection problems. A search ranking may return vivid pages with detailed language while missing terse administrative records. A model cannot correct absent evidence by writing a more balanced conclusion. The data pipeline needs coverage checks and a transparent statement about inaccessible or unmatched records.

AI can help compute budgets, normalize categories and identify inconsistent treatment, but it should not silently decide which people’s needs deserve attention. Sensitive allocations require accountable human governance and appropriate privacy protections. The purpose of using intelligence tools is to strengthen a fair evidence route, not to make an old visibility imbalance sound technically objective.

Studies on donation framing are not direct instructions for medical triage, disaster response or individual safeguarding. Such fields have legal, professional and ethical obligations that must override simplistic cost-per-person rankings. A person facing immediate danger should not be denied protection merely because another intervention serves a larger group on a spreadsheet.

Some outcomes cannot be measured precisely, and some populations have strong reasons to remain anonymous. Avoid designing eligibility processes that require disclosure of painful personal stories, public identification or images. Representation can empower people when it is voluntary and accurate; it can also exploit vulnerability when consent is weak or support depends on performance of suffering.

The research evidence does not support a universal numerical correction factor for identifiability. Do not multiply donations by a made-up bias coefficient or claim that a “debiasing exercise” is guaranteed to change behaviour. Treat framing as one possible influence and test interventions in the actual context where feasible.

Finally, beware the opposite extreme: treating all personal testimony as irrelevant noise. A case may reveal a flaw in policy, an unmeasured outcome or an urgent individual right. Intelligent institutions keep channels open for specific evidence while ensuring that people without compelling public narratives are not forgotten. Fairness requires both attentive listening and consistent comparison.

Frequently asked questions

Does the identifiable victim effect mean people are selfish? No. It concerns possible differences in attention or responses under certain forms of presentation, not a moral diagnosis of an individual.

Is the effect scientifically settled? No. Classic studies reported it, while a preregistered replication and reanalysis found no support in its tested conditions. General claims should be qualified by design and context.

Should charities stop telling personal stories? Not necessarily. Consented stories can communicate lived experience well. They should be paired with accurate programme-level evidence and transparent funding rules.

Do statistics make a decision automatically fair? No. Statistics can omit excluded groups, use misleading denominators or conceal important values. A fair process also requires legitimate goals, rights, privacy and accountability.

Is helping one named person inherently a bad allocation? No. Urgency, vulnerability, obligation and feasibility can justify individual support. The question is whether these criteria, rather than identity alone, explain the priority.

How do I compare one case with many? State a common decision target, measure feasible effects using explicit denominators and units, acknowledge uncertainty, and make ethical constraints visible.

What if a real story reveals information missing from the data? Include and verify that information, then look for the same missing factor in comparable cases. An anecdote can generate a valid hypothesis.

Can AI solve this bias automatically? No. AI summaries and ranking can reproduce selection and presentation effects. Human oversight, source coverage and explicit allocation rules remain necessary.

Source evidence and further reading

The original research by Small, Loewenstein and Slovic, “Sympathy and Callousness: The Impact of Deliberative Thought on Donations to Identifiable and Statistical Victims,” is at https://doi.org/10.1016/j.obhdp.2006.01.005 . Paul Slovic’s 2007 discussion of psychic numbing is at https://doi.org/10.1017/S1930297500000061 . The 2023 Collabra: Psychology preregistered replication and extension by Maier, Wong and Feldman is at https://doi.org/10.1525/collabra.90203 and the publisher’s article page is https://online.ucpress.edu/collabra/article/9/1/90203/199223/Revisiting-and-Rethinking-the-Identifiable-Victim . These sources should be read together rather than using one abstract as proof that all settings behave alike.

For related but different reader problems, see eduKateSG’s Availability Heuristic for vivid-memory effects on frequency and risk estimates, Denominator Neglect for problems caused by losing the relevant base count, and Hidden Profiles for group decisions that overemphasize shared information. The broader How Intelligence Works hub connects these mechanisms.

All scenarios and numbers in this article are fictional teaching examples. No claim is made that any school, family or charity in the examples experienced the reported outcomes. The recurring practice is to let individual accounts improve understanding while ensuring that evidence, rights and allocation rules remain visible for everyone affected.

Discover more from eduKate Singapore

Subscribe now to keep reading and get access to the full archive.

Continue reading