Analytic induction is a qualitative research method that uses inductive, as opposed to deductive reasoning. The process to be explained (explandum) and the factors that explain the phenomenon (explanans) are progressively redefined in an iterative process in order to maintain a perfect (universal) relationship between them.
By using an iterative process, a series of hypotheses are created upon examining cases, and as new cases are examined, the hypotheses are modified in such a way that cases are either consistent with the explanations or are placed outside the domain of the study. Each case is individually examined against the hypothesis, in order to validate or reject the assumption.
Analytic induction was first described in 1934 by Florian Znaniecki . Donald Cressey subsequently summarized Znaniecki’s rather complex description of AI by defining six steps . These are:
- a phenomenon is defined in a tentative manner
- a hypothesis is developed about it
- a single instance is considered to determine if the hypothesis is confirmed
- if the hypothesis fails to be confirmed either the phenomenon is redefined or the hypothesis is revised so as to include the instance examined
- additional cases are examined and, if the new hypothesis is repeatedly confirmed, some degree of certainty about the hypothesis results
- each negative case requires that the hypothesis be reformulated until there are no exceptions
Analytic induction involves using the concept of iterative hypothesis testing. The data (case set) is examined, and specific hypotheses are developed, which are tested against other cases. The hypothesis is validated or rejected in an iterative fashion. Initial AI studies stended to follow a stepwise approach, while AI studies in the 1970s began to have an ethnographic context.
- inductive rather than deductive
- well suited from an ethnographic viewpoint (good fit with the ethnographic narrative)
- orients observations towards developing a better hypothesis by allowing revision 
- allows redefining what is studied to better exclude exceptions
- initial attempts at universalism could not be realized
- Similar to grounded theory but has significant differences -- AI generates and tests theory provisionally, focuses on causation, requires all data to be tested against the hypothesis. Grounded theory focus more on generation and constant comparison 
Example of the Analytic Induction process
Stages in the analysis of field notes in a qualitative study of ear, nose, and throat surgeons' disposal decisions for children referred for possible tonsillectomy and adenoidectomy (with examples)11:
- Provisional classification—for each surgeon all cases categorized according to disposal category used (tonsillectomy and adenoidectomy or adenoidectomy alone)
- Identification of features of provisional cases—common features of cases in each disposal category identified (most tonsillectomy and adenoidectomy cases found to have three main clinical signs)
- Scrutiny of deviant cases—include in (2) or modify (1) to accommodate deviant cases (tonsillectomy and adenoidectomy performed when only two of three signs present)
- Identification of shared features of cases—features common to other disposal categories (history of several episodes of tonsillitis)
- Derivation of surgeons' decision rules—from the features common to cases (case history more important than physical examination)
- Derivation of surgeons' search procedures (for each decision rule)—the particular clinical signs looked for by each surgeon
Repeat steps (2) to (6) for each disposal category
from Bloor M. On the analysis of observational data: a discussion of the worth and uses of inductive techniques and respondent validation. Sociology. 1978;12:545–552. quoted in Pope C, Ziebland S, Mays N. Qualitative research in health care: Analysing qualitative data. BMJ. 2000 January 8; 320(7227): 114–116.
EXAMPLES IN INFORMATICS
Rogers M, Todd C. Information exchange in oncology outpatient clinics: source, valence and uncertainty. Psychooncology. 2002 Jul-Aug;11(4):336-45.
Qualitative study that explored the exchange of cancer information between physicians and patients -- transcripts were analysed using analytic induction and microinteractional analysis techniques.
Banet GA, Jeffe DB, Williams JA, Asaro PV. Effects of implementing computerized practitioner order entry and nursing documentation on nursing workflow in an emergency department. J Healthc Inf Manag. 2006 Spring;20(2):45-54.
This study looked at the effects of implementing CPOE and nursing documentation in an ED -- analytic induction was used to evaluate responses and categorize them into themes
- Znaniecki, F (1934). The method of sociology. New York: Farrar & Rinehart.
- Cressey, D R (1953) Other People's Money. Free Press, Glencoe
- Robinson, W. S. (1951). The logical structure of analytic induction. American Sociological Review, 16, 812-818.
- Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory. New York: Aldine.