A model of information seeking that characterizes search as an iterative, opportunistic and nonlinear process in which users collect discrete pieces of information from multiple sources over time, continually reformulate queries, follow leads (citations, hyperlinks, references) and combine partial answers rather than executing a single, optimal query that yields a complete result.

Principle

Information needs are often satisficed progressively: users adapt strategies based on intermediate findings, so retrieval systems should support iterative exploration, easy pivoting, provenance tracing and incremental integration of results rather than optimizing only for one‑shot precision per query.

Demonstration

Illustrative scenario — Situation: A student researching urban heat islands begins with a news article, follows cited studies to academic databases, saves a useful dataset from a municipal portal, refines search terms after each discovery and assembles a literature map. Recognition: The student recognizes that useful evidence arrives piecemeal and from varied formats. Action: She adapts queries and tools, records sources and integrates snippets into an annotated bibliography. Consequence: The final synthesis emerges from multiple short interactions rather than a single comprehensive search.

Misapplication

Error: modeling information seeking as a single linear pipeline (query → ranked results → selection) and designing interfaces exclusively for that flow. Why plausible: classical IR evaluation frames retrieval as one‑shot relevance. Semantic error: ignores user adaptation, serendipity and cross‑source integration. Correct interpretation: search interaction is dynamic and often requires support for incremental, heterogeneous gathering.

Consequence

Acknowledging berrypicking shifts system design and evaluation toward features that support session continuity, bookmarking, cross‑source linking, query histories and mixed‑format retrieval; it also informs user education about search strategies and the limits of precision metrics.

Reversal

For constrained tasks with a narrowly defined information need (e.g., known‑item retrieval, transactional queries or strictly specified metadata lookups), one‑shot query models perform adequately and the full berrypicking interaction pattern may be unnecessary.

Boundary

Clearly within: exploratory research, literature reviews, investigative reporting and learning tasks where the information need evolves. Boundary case: systematic review protocols that structure iterative search but require exhaustive reproducibility. Clearly outside: transactional lookups (e.g., a phone number) where a single precise query suffices.

Semantic Tension

Tension between supporting opportunistic, exploratory user behavior (flexible interfaces, serendipity) and optimizing algorithmic measures of single‑query precision and recall; tension between long‑session support and resource constraints on system complexity.

Synthesis

Berrypicking reframes search as a user‑driven assemblage process: effective retrieval systems and evaluation frameworks should privilege interaction continuity, provenance and tools for piecemeal synthesis rather than treating each query as an isolated retrieval event.