ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam...

20
ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London www.dylanwiliam.net

Transcript of ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam...

Page 1: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

ETS Confidential & Proprietary

Diagnostic testing that just might make a difference

Dylan Wiliam

Institute of Education, University of London

www.dylanwiliam.net

Page 2: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

2

Diagnostic Items in Mathematics and Science: Project rationale• Traditional testing deals with individuals, but teachers

mostly deal with groups• Data-Push vs. Decision-Pull

– Data-push• Quality control at end of an instructional sequence• Monitoring assessment• Identifies that remediation is required, but not what• Requires new routines to utilize the information

– Decision-Pull• Starts with the decisions teacher make daily• Supports teachers “on-the-fly” decisions

• If a 30-item test provides useful information on an individual, then responses from 30 individuals on a single item might provide useful information on a class

Page 3: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

3

Premises of the DIMS project

• Single, well-developed, questions provide a reasonably strong basis for real-time instructional decision-making

• Questions with principled and interpretable incorrect answers can be used to elicit evidence of student thinking

• Such questions, when used with all-student response systems, and when followed with discussion, can advance student learning

• Such questions can deepen teacher knowledge (both content and content-pedagogical)

Page 4: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

4

The DIMS project in practice

• Focuses on 4th & 8th grade, mathematics & science

• Includes a bank of approximately 150 diagnostic questions per subject/grade level

• Question bank is a complementary resource for any curriculum, not a curriculum replacement

• Items to be used one at a time, within the flow of day-to-day lessons

• Focuses on collecting evidence about student learning in order to adapt instruction to meet the learning needs of students in real time

Page 5: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

5

Item development

• Review of state standards• Review of relevant literature on student

conceptions• Identification of significant (mis-)conceptions• Item construction

– appears to be very difficult for many item-writers– a distractor-stem-key approach worked for some

• Item review and editing– Expert review– Piloting– Final review and editing

Page 6: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

6

DIMS meets traditional psychometrics

• Concept-based distractors increase difficulty

• Most IRT models assume– Monotonicity

– Equivalence of incorrect responses

• Need for analysis of single items makes most theories difficult to apply, or irrelevant

Page 7: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

7

Semi-dense items(Bart, Post, Lesh & Behr, 1994)

Five properties

• Response interpretability

• Response discrimination

• Rule discrimination

• Exhaustive set usage

• Semi-density

Page 8: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

8

Response interpretability

Cognitive Rules Responses

A

B

C

D

Each response interpretable by at least one cognitive rule

Page 9: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

9

Response discrimination

Cognitive Rules Responses

A

B

C

D

Each response interpretable by exactly one cognitive rule

Page 10: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

10

Rule discrimination I

Cognitive Rules Responses

A

B

C

D

Response discrimination + uniqueness of cognitive rules that interpret a response

Page 11: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

11

Exhaustive rule set usage

Cognitive Rules Responses

A

B

C

D

Response discrimination + every cognitive rule interprets at least one response to the item.

Page 12: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

12

Semi-density

Cognitive Rules Responses

A

B

C

D

Exactly one cognitive rule interprets each response to the item and each cognitive rule interprets exactly one response

Page 13: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

13

Discriminating incorrect cognitive rules (Hart, 1981)

Version 1

If e+f = 8, then e+f+g = 

• 9

• 12

• 15

• 8+g

Version 2

If f+g = 8, then f+g+h = 

• 9

• 12

• 15

• 16

• 8+h

Page 14: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

14

Discriminating correct cognitive rules (Bart et al., 1994)

Version 1

a) 12

b) 24

Version 2

a) 24 cents, because 8×3=24 or 8 pieces × 3 cents per piece=24 cents.

b) 24 cents, because 4×6=24. c) 24 cents, because 2/6=8/x and 2x=48 for which

x=24.d) 24 cents, because 2/6=8/x and 2/6x4/4=8/24.e) 24 cents, because 2/6=4/12=6/18=8/24.f) 24 cents, became 2/6=4/12=8/24.g) 12 cents, because 2+4=6 implies that 8+4=12

Ann and Kathy each bought the same kind of bubble gum at the same store. Ann bought two pieces of gum for six cents. If Kathy bought eight pieces of gum, how much did she pay?

Page 15: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

15

Discriminating between incorrect and correct cognitive rules

Version 1

There are two flights per day from Newtown to Oldtown. The first flight leaves Newtown each day at 9:20 and arrives in Oldtown at 10:55. The second flight from Newtown leaves at 2:15. At what time does the second flight arrive in Oldtown? Show your work.

Version 2

There are two flights per day from Newtown to Oldtown. The first flight leaves Newtown each day at 9:05 and arrives in Oldtown at 10:55. The second flight from Newtown leaves at 2:15. At what time does the second flight arrive in Oldtown? Show your work.

Page 16: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

16

Discriminating between incorrect and correct cognitive rules (2)

B C DA

B C DA

Page 17: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

17

Conclusion

Cognitive Rules Responses

A

B

C

D

Correct

Incorrect

Page 18: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

18

Conclusion (2)

• For an item to support instructional decision-making, the key requirement is that in no case do incorrect and correct cognitive rules map on to the same response.

• If this property is met, then the semi-density properties are less important.

• If this property is not met, then the semi-density properties are irrelevant.

• The discovery of new incorrect cognitive rules that interpret item keys leads to item improvement

Page 19: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

19

Conclusion (3)

• Data on impact on student achievement not yet available

• Evidence of impact on– Teachers’ practice

– Student engagement

– Student affect

Page 20: ETS Confidential & Proprietary Diagnostic testing that just might make a difference Dylan Wiliam Institute of Education, University of London .

20

Questions?

Comments?