Engaging students outside of class as a tool for enhanced learning Dan Bernstein University of Kansas UBC Centre for Teaching and Academic Growth 22 November 2006 djb@ku.edu Unhappy with answers to questions • What is the major point of quantum mechanics that makes prediction impossible? What level of prediction is affected? What level is not? • What effect do quantum mechanics have on scientific objectivity as part of our process of knowing? • Describe the Heisenberg uncertainty principle (briefly). What relevance do you think it has for contemporary research and clinical practice in psychology? • Does quantum theory render Newtonian mechanics useless to humans in their interactions with the physical environment? Why or why not? Simple problem led me onto a path of continuous inquiry that has lasted a decade Started with Experimental Studies • In search of the best way to generate high level student answers to those questions • Compared multi-media presentation with lecture, reading, and irrelevant information • Measured rote and complex understanding with multiple choice, fill-in, short essay • Advantages of random assignment and controlled conditions • Goal ultimately a practical one Basic Procedures • Commercial and home-made multi-media • Used video and live lecture with same content as multi-media • Created text samples with same content • Assessment of levels of understanding with mixed methods of responding • Individual instruction with unlimited time • Participants from introductory psychology Learning from Lecture 40 35 30 25 20 Lecture 15 10 5 0 <= 69 70-79 80-89 90-100 Learning from Lecture v. Reading 60 50 40 Lecture Readings 30 20 10 0 <= 69 70-79 80-89 90-100 Lecture, Reading, and Computer 60 50 40 Lecture Readings Computer 30 20 10 0 <= 69 70-79 80-89 90-100 Percentage Correct by Instructional Mode and Level of Understanding 70 60 50 40 Rote Application Evaluative* 30 20 10 0 Computer Lecture Readings Percentage Correct by Instructional Mode and Question Type 80 70 60 50 Multiple-choice* Fill-in-the-blank Essay 40 30 20 10 0 Computer Lecture Readings Mean Instructional Time by Instructional Mode 30 25 20 15 Instructional time* 10 5 0 Computer Lecture Readings Upgraded the lecture component • Live lecture to portion of introductory class • Topic is reactive measurement and quantum mechanics • Matched to locally authored multi-media and text -- also irrelevant text control • Motivation sustained by possibility of test • Assessment layered by type of understanding and format of question Distribution of total test performance resulting from four methods of instruction Number of Correct Responses Across Levels of Understanding 14 12 10 lecture computer reading control 8 6 4 2 0 rote application evaluation Time on task (in minutes) 70 60 50 40 30 20 10 0 Computer Lecture Reading What do you make of these data? • How would it help me teach? • Any ideas for next innovation? • Critical reactions to the evidence? My take home messages • Students learning by computer or reading showed greater and deeper understanding • Multi-media has potential but very high cost to get started • Reading consistently took the least time • Reading is a good way to learn if you have a way to assure that it is done Next problem arose when I presented my data • Challenged by Tom Hatch, from Gardner’s Project Zero group • Described “Deep Understanding” – Use ideas in a context not taught – Ideas retained without memorization • I presumed that abstract was more general • Agreed to test the idea in my class Tracking understanding over time • • • • • • • • • Learning and motivation Sophomore level course 20 to 45 students Taught 9 times in 7 years Professional grade reading Hybrid online and F2F Largely discussion in class All essay examinations Conceptual writing required Wanted Richer Essay Exams • Why is magazine training so critical to the process of shaping a new response class? Sometimes the sound of the operation of the food magazine is called a bridging stimulus; what might this metaphor mean? • What is the evolutionary mechanism for the cultural survival of religious taboos against slaughtering and eating cattle? You should identify the conditions that make this evolution most likely. • Explain the notion of selection by consequences as a description of the change in distribution of behavior that takes place during shaping. How is this process different from selection that takes place under naturally occurring conditions? Reasonably Conceptual Items • Describe the results of [Razran’s] second experiment on semantically mediated generalization (numbers) done with a 13-year old boy. Be sure to identify the respondent behavior, the training stimuli, and the test stimuli. Why is this example semantically rather than physically mediated generalization? • What do chicken scratches, raccoon rubbings and pig rooting have in common? What stimuli have pre-existing relations these bits of behavior? What does the “misbehavior of organisms” have to do with the generality of principles of learning? Transform into Problems • Old question: – What were the reinforcing consequences in the Welsh, Bernstein & Luthans (restaurant) study? How were the consequences identified? • Problem-based Assessment : – Suppose you were asked to implement a motivational program in a business with 25 employees engaged in the following activities: production planning, inventory delivery, direct production, packaging, and marketing. Based on your understanding of the restaurant study, how would you improve the quality of the employees work by using access to activities as a motivator? Describe the costs and benefits of the program and make a recommendation about implementation. • Old question: – The behavioral context into which punishment is added must be considered in evaluating the likely effect of a given stimulus. What conditions are essential in this analysis? • New question: – Suppose it is your task to decrease the frequency of college students engaging in binge drinking or games involving high levels of intoxication. You have been asked to set up a punishment program to eliminate this program before someone is hurt. What context for binge drinking would you identify first? What punishing consequence would you use? What characteristics would you include in your punishing system to maximize effectiveness? Please give an example of an additional element you would need to include to make the plan maximally effective. Points for Question 1 • 4pts - reinforcement for drinking • 5 pts - ID specific severe punishing consequence • 4 pts - Practical plan for measurement of abusive drinking • 12 pts - Immediate consequence (4 pts); severity (more than change action) (4 pts); consistent system (4pts) • 5 pts - reinforcement plan for alternative actions that are preferred, if possible using original class of reinforcers • Old question: • Describe the results of the experiment on semantically mediated generalization done with a 13-year-old Soviet boy. Be sure to identify the respondent behavior, the training stimuli, and the test stimuli. Why is this example semantically rather than physically mediated generalization? • New question: • Generate your own example of classical conditioning using the neutral stimuli “idea” and “rock” along with the eliciting relation between a very loud noise and increased heart rate. Your example should include all of the following components: a description of a conditioning procedure that would produce differing reactions to the two stimuli, a description of the procedure that tests for the direct effects of conditioning, a description of a procedure that would test for physically mediated generalization, a description of a procedure that would test for semantically mediated generalization, and the likely results of the three test procedures. Points for Question 2 • 4 pts - specific conditioning -- say “idea” (2) and then puff of air in eye (2) • 7 pts - Direct test: Present “idea” (3) w/o air (2) and see blink (2) follows • 7 pts - Physically mediated generalization: Present sound similar to “idea” (3) w/o air (2) and see weak or little blink (2) • 7 pts - Semantically mediated generalization: Present sound or picture or word related to “idea” (3) w/o air (2) and see strong blink (2) more than physically mediated Use existing work: Keep an archive of evidence Distribution on 1st Attempt Old questions 40 35 30 25 Percent 20 Students 15 10 5 0 <= 69 70-79 80-89 Levels of Achievement 90-100 Change of Questions Old questions New questions 50 45 40 35 Percent 30 25 Students 20 15 10 5 0 <= 69 70-79 80-89 Levels of Achievement 90-100 What do you make of the data? • Is this a surprise? • What examples do you have in your own work? • What does it take to “problematize” a concept for assessment? • Keep going in this direction or turn back to abstraction? I relied on my experimental results • The key is getting students to read • If only they were prepared, we would all be better off • Post hoc argument from fluency -free up mental resources for problem solving [away from details] Organizing units by topics Online questions on reading Feedback until mastery of topics Students consider reading Complete roster services Click on name for full record See items, make comments Aggregate Performance 1st Try New Questions 35 30 25 Percent 20 Students 15 10 5 0 <=69 70-79 80-89 Levels of Achievement 90-100 Addition of EDU Out of Class New Questions Web-aided 40 35 30 25 Percent 20 Students 15 10 5 0 <=69 70-79 80-89 Levels of Achievement 90-100 What do you make of these data? Would you keep using it? I did. Next idea came from Grant Wiggins • • • • Welding teacher story Generate internalized criteria Like composition studies and editing Used prior student work with permission • No feedback from me Bb - consider others’ work Students share reactions w/o me Exam - Eliminate Problem Actions No Web W/ Examples 70 60 50 Percent 40 Students 30 20 10 0 (0-3) (4-6) (7-9) (10-12) Categories of Achievement (13-15) Exam Question on Cultural Anthropology No Web W/ Examples 80 70 60 50 Percent 40 Students 30 20 10 0 (0-4) (5-8) (9-12) (13-16) (17-20) Categories of Achievement Take home complex essay Without With Examples 70 60 50 Percent 40 Students 30 20 10 0 (0-5) (6-10) (11-15) (16-20) Categories of Achievement (21-25) Add Generation to Recognition • • • • • Posted another variation on the question Each student was to write an answer Also expected to comment on one answer Collected a sample of answers They were focus of class discussion – Students comment on completeness – Offer suggestions to refine answers Learning With Full Techno Boost Report: http://www.unl.edu/peerrev/examples/bernstein/index.html Old questions New questions Most recent 50 45 40 35 Percent 30 25 Students 20 15 10 5 0 <= 69 70-79 80-89 Levels of Achievement 90-100 Comparison with last three offerings (percent of students) 60 50 40 Fall 00 Fall 01 Spring 03 Spring 04 30 20 10 0 <=69 70-79 80-89 90-100 Distribution of Performance Median 80/90 8 7 6 5 spring 03 spring 05 4 3 2 1 0 <=69 70-79 80-89 90-100 General Pattern of Improvement • Increased frequency and care of student reading • Dramatically improved class climate • Increased recognition of examples of understanding • Improved percentage of students reaching high levels of understanding Note use of equivalent problems • Same idea with different contextual cues • Generate up to eight problems per concept • Use several to teach, then test in new context • Teach again as needed in new examples • Retest in yet additional new context • Final exam is yet another generalization test Inductive and deductive teaching • • • • • Active discovery is important learning Didactic instruction can be efficient Bransford -- “time for telling” My research => advantages for both Lendol Calder is a nice synthesis Comments? Examples? Questions?