Analogies Between Science and Design: What Models of Science Can Learn from Models of Engineering Design? Christian Schunn University of Pittsburgh
What I won t talk about Psychology of Science Complex Spatial Cognition Science Education
Scientific Discovery & Engineering Design Competing Epistemologies? SCIENCE: ANALYTIC GOAL: EXPLANATION OF NATURAL PHENOMENA GENERATE HYPOTHESIS DESIGN EXPERIMENT INTERPRET DATA CONDUCT EXPERIMENT REQUIREMENTS IDEATION CONCEPT MATRIX PROTOTYPE GENERATION PROTOTYPE TESTING ANALYSIS OPTIMIZATION ENGINEERING: SYNTHETIC GOAL: CREATION OF ARTIFACTS TO MEET NEEDS
Not to be confused by people doing both Bohr Pasteur BASIC (UNDERSTAND) YES NO Edison STOKES (1997) NO APPLIED (BUILD) YES
Scientific Discovery & Engineering Design Twins separated at birth? SCIENCE CONFIRMATION BIAS ERROR MINIMIZATION HEURISTIC SEARCH ANALOGY DIVERGENT PROCESS CONVERGENT PROCESS EMPIRICAL TESTING DESIGN FIXATION ANALYSIS + OPTIMIZATION ENGINEERING
More mappings between Science & Engr. Design FIND REGULARITY => DESIGN HYPOTHESIS HYPOTHESIS =? PROTOTYPE EXPERIMENT =? PROTOTYPE MODEL =? MODEL => DESIGN EXPERIMENT => DESIGN INSTRUMENT DESIGN SOLUTION => ANALYZE DATA => EXPLAIN FAILURES => DEVELOP MODELS
AI of Engineering & Discovery: Different Social Worlds? Read papers of conference attendees (who building AI models) Anyone cite work in engineering design? Not except Ashok Goel!
Review of Engr. Design Literature: What Works Verbal + Other 23% Verbal Protocol Only 23% Ethnography 10% Interviews 8% Video protocol 5% Written protocol 5% Case Study 27% MEHALIK, M. M., & SCHUNN, C. D. (2006). WHAT CONSTITUTES GOOD DESIGN? A REVIEW OF EMPIRICAL STUDIES OF THE DESIGN PROCESS. INTERNATIONAL JOURNAL OF ENGINEERING EDUCATION, 22(3).
Review of Engr. Design Literature: What Works Search the space (explore alternatives) 95% mention, 61% good Use interactive/iterative design methodology 93% mention, 51% good Explore problem representation 90% mention, 69% good Explore scope of constraints 55% mention, 59% good Redefine constraints 45% mention, 56% good MEHALIK, M. M., & SCHUNN, C. D. (2006). WHAT CONSTITUTES GOOD DESIGN? A REVIEW OF EMPIRICAL STUDIES OF THE DESIGN PROCESS. INTERNATIONAL JOURNAL OF ENGINEERING EDUCATION, 22(3). }NEW(ISH)?
The Environment of Analogical Reasoning in Expert Conceptual Design CHRISTENSEN, B. T., & SCHUNN, C. D. (2007). THE RELATIONSHIP OF ANALOGICAL DISTANCE TO ANALOGICAL FUNCTION AND PRE-INVENTIVE STRUCTURE: THE CASE OF ENGINEERING DESIGN. MEMORY & COGNITION, 35(1), 29-38.
Studied Setting Followed 5-person product development team in a major international company dealing in medical plastics design won many design awards; industry leader Videotaped scheduled weekly brainstorming and product development meetings Analyzed 7 product development meetings (9 hours of video), collected over 5 months (conceptual design phase) 5800 segments of on-task speech coded
Research Question 1 What kind of analogies do highly innovative designers use? Within-domain analogies? (More relevant) Between-domain analogies? (More novel)
Distribution of analogies (N=102, 1 every 5 minutes) Within 55% Between 45% POTATO PRINT ZIPPERS CREDIT CARDS MILK CARTONS SHOES CARS TOILET PAPER CHRISTMAS DECORATIONS WATER WHEELS PICTURE PUZZLE VENETIAN BLINDS UNDERWEAR
Research Question 2 What Role Do Analogies Play in Design? Just about explanation? Dunbar (1995) claimed long distance analogies only after the fact for explanation New answers? Or maybe potential problems?
Observed Functions of Analogies in Design Explain to others 32% Solve Problem 40% Identify Problem 28%
Distance by Function 100 % BETWEEN 75 50 25 0 Identify Problem Solve Problem Explain Identify are local, explain are long distance Solve problem analogies are mixed! (different from science?)
Research Question 3 How does the external design environment shape what analogies occur (inspiration vs. design fixation)? no support (unconstrained) sketches (loosely constrained) prototypes (tightly constrained) 96% OF DESIGNERS USE SKETCHES OR PROTOTYPES DURING SOLUTION DEVELOPMENT
Frequency of analogies by external environment 5 4 % OF SEGMENTS 3 2 1 0 Idea Sketch Prototype Fewer Analogies w/ Prototypes (NO CONFOUND W/ PHASE OF DESIGN)
Interaction with distance 5 % of segments 4 3 2 1 Idea Sketch Prototype 0 Within-domain Between-Domain Prototype drop in analogies specific to betweendomain analogies
So what? Sketches likely help because (according to Barbara Tversky) they are ambiguous (re-interpretable, generative)
More things Science could learn from Engineering Design Fixation is a real pain Abstraction (like ambiguity) is your friend Relevant stimulation is the key to incubation