Methods & Data in Human Sciences (DP IB Theory of Knowledge): Revision Note

Methods & data in human sciences

• The Human Sciences use the scientific method to acquire knowledge

• Knowledge claims in the human sciences depend on:

  • how methods shape what counts as evidence

  • how that evidence is interpreted

• Disagreement often comes from:

  • different perspectives

  • different questions asked

  • different data selected

  • different interpretations

  • different standards of justification

Observations, surveys and interviews

Observations

• Observation is somewhat problematic as we observe the outward manifestations of human behaviour, but cannot consistently observe the inner workings

• During an observation, researchers watch and record people’s behaviour to identify patterns and meanings, without deliberately changing variables as they would in an experiment

• The act of being observed may affect behaviour and impact the reliability of findings

• Observations prioritise behaviour in context

• Barriers to observation include linguistic and cultural differences

• Participant observation is an important method, especially in anthropology

• Observational data is likely to produce correlation data rather than data that demonstrates that changing one variable causes a change in another

  • This means justification tends to be probabilistic rather than certain

Surveys

• During a survey, researchers ask a standard set of questions to many people

• This allows for the collection of comparable data about attitudes, beliefs or behaviours

• Survey data can be limited because:

  • people may misunderstand questions, guess or give socially acceptable answers rather than truthful ones

  • surveys can also miss important context because fixed questions simplify complex experiences

    • E.g. a wellbeing scale compresses feelings into numbers; this data may not explain why people answered as they did

  • survey wording frames what counts as an answer, so it can introduce bias,

    • E.g. asking “How harmful is social media?” presupposes harm; this increases negative ratings and strengthens a harm claim without new behaviour evidence

Interviews

• In an interview, a researcher asks a participant questions in a conversation to gather detailed accounts of experiences, beliefs and reasons behind behaviour

• Interviews provide evidence about reasons and meanings

  • This supports interpretive knowledge claims

• However, interview dynamics may shape interviewee responses

  • Power relations and trust affect what is disclosed

  • Missing or filtered testimony limits what can be justified

    • E.g. a student interviewed by a teacher withholds criticism, so the evidence supports “high satisfaction” more than it should

Quantitative vs qualitative methods

Quantitative methods

• Quantitative methods support pattern claims

  • Aggregation (combining many individual data points into a summary measure (e.g. totals, averages, percentages) can justify general claims about trends and relationships

  • This does not automatically justify claims about individuals, e.g.:

    • A large dataset links study time and grades

    • This supports a correlation claim but not a single-cause claim for any one student

• Quantitative results can seem more authoritative than they are

  • Precision in numbers can be mistaken for certainty in interpretation

  • This can inflate the strength of the knowledge claim beyond the evidence

Qualitative methods

• Qualitative methods support context-sensitive meaning claims

  • Rich detail can justify claims about how people understand experiences in a setting

• Qualitative evidence increases interpretive flexibility

  • This means that multiple plausible interpretations can fit the same data

  • Disagreement can be about standards of justification, not just the data itself

Method choice

• Method choice reflects the knowledge aim

  • Prediction and comparison often prioritise quantitative evidence

  • Understanding meanings often prioritises qualitative evidence

  • Each uses different criteria for what counts as good justification

• Mixed methods can strengthen justification

  • Quantitative patterns can be paired with qualitative explanations

  • This can reduce uncertainty about what the pattern means, e.g.:

    • Survey trends suggest rising anxiety

    • Interviews clarify plausible sources, strengthening interpretation

Modelling and prediction

• A model is a simplified representation linking variables to outcomes

  • Simplification can increase predictive usefulness

  • However, simplification can also reduce explanatory completeness, e.g.:

    • A model predicts absenteeism from commute time and grades

    • This supports prediction claims but not a full explanation of why absence occurs

• Predictions depend on the stability of the context

  • Changing norms or incentives can weaken future reliability

  • Past fit does not guarantee future justification

• Predictions in the human sciences are often probabilistic

  • Even a strong model supports likelihood claims rather than certainty

Data reliability and interpretation

• Reliability problems can weaken justification

  • If measurements are inconsistent, the evidence base is unstable

  • Conclusions become fragile under remeasurement, e.g.:

    • Different observers code the same behaviour differently

    • A claimed pattern disappears across repeats

• Reliability is not the same as validity

  • Data can be consistent while measuring the wrong construct

  • This supports confident claims that are not well-grounded, e.g.:

    • A test score is measured consistently and treated as “intelligence”

    • The claim exceeds what the data justifies

• Correlation is vulnerable to overinterpretation

  • Patterns can be mistaken for causal claims

  • Causal language needs extra justification beyond association

• Data cleaning decisions shape conclusions

  • Handling outliers and missing data can change patterns

  • Changed patterns change what interpretations are justified, e.g.:

    • Removing “extreme” responses raises average satisfaction

    • The more positive claim comes from method choices, not necessarily a changed reality