📄 Analogical problem solving

Note for the library: The provided PDF extraction contains only page markers (pp. 1–50) and no text. The summary below is based on the paper's widely known content, not on the supplied file. Exact percentages and the experiment count are omitted because they could not be checked against the text. Please re-extract the PDF if precise figures are needed.

Summary

Gick and Holyoak study how people solve a new problem by drawing on a structurally similar situation from a different domain. Their target task is Duncker's classic radiation problem. A doctor must destroy an inoperable tumor with rays, but rays strong enough to kill it would also destroy the healthy tissue they pass through. In a series of experiments, participants first read a story that served as a source analog. In the main story, a general cannot send his whole army down any single mined road to a fortress, so he splits it into small groups that converge from several directions. Participants then attempted the radiation problem, sometimes thinking aloud. The authors measured how often the analogous "convergence" solution appeared, with and without a hint to use the story. People could map the story onto the problem and transfer its solution, but they often failed to notice the connection unless prompted. The work established a mapping-based account of analogical transfer and became a cornerstone of research on analogy in cognition and AI.

Key points

• Target problem: Duncker's radiation problem. The analogous solution is several weak rays aimed from different angles that meet at the tumor.
• Source analogs were military stories, chiefly the fortress-attack ("General") story. Variants with different solutions led participants toward correspondingly different answers, which shows that the specific solution structure transferred.
• Without the source story, convergence solutions were rare. Reading the story helped, but the largest gains came when participants were explicitly told to use it.
• The core finding is that retrieving or noticing an analogy is a major bottleneck. Once the link is pointed out, mapping and applying the solution is comparatively easy.
• Analogical problem solving is framed as mapping correspondences between a source and a target representation. Think-aloud protocols showed that this mapping was often partial.