The Fut ur e of Assessment How wi l l AI, Aut omat i on, and Mac hi ne Lear ni ng Change How We Devel op and Del i ver Assessment s?

Similar documents
Fr anchi s ee appl i cat i on for m

P a g e 5 1 of R e p o r t P B 4 / 0 9

Senility Degree. Our machine derives APG waveform after 2 nd differential of arterial pulse wave in order to measure

Animals and Behaviors. Templeton Biology

T h e C S E T I P r o j e c t

P a g e 3 6 of R e p o r t P B 4 / 0 9

OH BOY! Story. N a r r a t iv e a n d o bj e c t s th ea t e r Fo r a l l a g e s, fr o m th e a ge of 9

Hybrid Bonded Wheel

A L A BA M A L A W R E V IE W

Gen ova/ Pavi a/ Ro ma Ti m i ng Count er st at Sep t. 2004

Alles Taylor & Duke, LLC Bob Wright, PE RECORD DRAWINGS. CPOW Mini-Ed Conf er ence Mar ch 27, 2015

176 5 t h Fl oo r. 337 P o ly me r Ma te ri al s

Class Discussions. The Glue Between Reading, Writing, and Understanding

Per cent Wor d Pr oblems

RAHAMA I NTEGRATED FARMS LI MI TED RC

The Nature of Engineering

THIS PAGE DECLASSIFIED IAW E

Human Anatomy - Brain

M Line Card Redundancy with Y-Cab l es Seamless Line Card Failover Solu t ion f or Line Card H ardw or Sof t w are Failu res are Leverages hardware Y-

Table of C on t en t s Global Campus 21 in N umbe r s R e g ional Capac it y D e v e lopme nt in E-L e ar ning Structure a n d C o m p o n en ts R ea

GROWMARK, INC 2200 Sout h Ave nue, P. O. Box 587, Counc i l Bl uf f s, I A

The Unjust Steward THE MAN WHO LOST HI S JOB

NEC and OSS NEC Co r p o r a t i o n 2007

WARNI NGLETTER CERTI FI ED MAI L RETURN RECEI PT REQUESTED. Ref er ence No. 06- HFD

I zm ir I nstiute of Technology CS Lecture Notes are based on the CS 101 notes at the University of I llinois at Urbana-Cham paign

Software Process Models there are many process model s in th e li t e ra t u re, s om e a r e prescriptions and some are descriptions you need to mode

Instruction Sheet COOL SERIES DUCT COOL LISTED H NK O. PR D C FE - Re ove r fro e c sed rea. I Page 1 Rev A

B2B Mi ddl ewar e Devel opment ( Sony)

I N THE COURT OF APPEALS OF TENNESSEE EASTERN SECTI ON

Fall / Winter Multi - Media Campaign

Provider Satisfaction

o Alphabet Recitation

H STO RY OF TH E SA NT

P-4( ?"1ST ttl 4 s r 6 e L. /tv NI/.cor GPAv-lb Lo U e. oufax,v 4 /9y, 993 -

TFCC / TCDP / TCPP / TCSA and Proposal f or a new TC on Scalable Comput ing (TCSC)

Agenda Rationale for ETG S eek ing I d eas ETG fram ew ork and res u lts 2

REQUEST FOR PROPOSAL N 75/ 2014

Physics 663. Par t icle Physics Phenomenology. May 7, Physics 663, lecture 8 1

I nt er nat i onal psychoanal ysi s. net was l aunched i n Januar y On t he

Use precise language and domain-specific vocabulary to inform about or explain the topic. CCSS.ELA-LITERACY.WHST D

graphicdesign SPECIAL INFORMATION & DISPENSER ORDER FORM Version1.0,Updated20October2016 Date SP-4066 Encore DEF Graphics Form Manual

I N A C O M P L E X W O R L D

Use precise language and domain-specific vocabulary to inform about or explain the topic. CCSS.ELA-LITERACY.WHST D

K owi g yourself is the begi i g of all wisdo.

, L.L.C. (Ma na g e r Ma na g ed) OPERATIN G AGREEMEN T

Soil Stabilization for Pavements

THIS PAGE DECLASSIFIED IAW EO IRIS u blic Record. Key I fo mation. Ma n: AIR MATERIEL COMM ND. Adm ni trative Mar ings.

Lesson Ten. What role does energy play in chemical reactions? Grade 8. Science. 90 minutes ENGLISH LANGUAGE ARTS

USER MANUAL V1.3 CHRYSLER VEHICLES VEHICLE FLASHER

CATAVASII LA NAȘTEREA DOMNULUI DUMNEZEU ȘI MÂNTUITORULUI NOSTRU, IISUS HRISTOS. CÂNTAREA I-A. Ήχος Πα. to os se e e na aș te e e slă ă ă vi i i i i

WEATHER MAP INFORMATION STATION MODEL. Station Model Lab. Period Date


Le classeur à tampons

Progression in calculations Years 5 and 6

I N THE COURT OF APPEALS OF TENNESSEE EASTERN SECTI ON

THE IMF THE IMF. The I nt er nat i onal Monet ar y NOT PERFECT, BUT ESSENT IA L. R o b e riat LD. H o r m a t s

Executive Committee and Officers ( )

J A D A V PUR U N IV ERS IT Y K O LK AT A Fa cu lty of En gi n eer in g & T e ch no lo gy N O T I C E

Building Harmony and Success

COMPILATION OF AUTOMATA FROM MORPHOLOGICAL TWO-LEVEL RULES

PUPI L PREMI UM POLI CY. June 2017

PC Based Thermal + Magnetic Trip Characterisitcs Test System for MCB

Foreword by Yvo de Boer Prefa ce a n d a c k n owledge m e n ts List of abbreviations

X2 DESIGNER W E D D I N G S

What are S M U s? SMU = Software Maintenance Upgrade Software patch del iv ery u nit wh ich once ins tal l ed and activ ated prov ides a point-fix for

Me n d e l s P e a s Exer c i se 1 - Par t 1

Peace Prevails School Programme Background

The Ind ian Mynah b ird is no t fro m Vanuat u. It w as b ro ug ht here fro m overseas and is now causing lo t s o f p ro b lem s.

AT LAST!! CAGE CODE 6CVS2. SandMaster 20 for Skid Steers THE FUTURE OF EMERGENCY FLOOD CONTROL HAS ARRIVED.

P rac t i c e plotting points in cartesian coordinate system.

I M P O R T A N T S A F E T Y I N S T R U C T I O N S W h e n u s i n g t h i s e l e c t r o n i c d e v i c e, b a s i c p r e c a u t i o n s s h o

COLLECTIVE AGREEMENT BETWEEN BFI CANADA INC. AND. December 5, December 4, DON McGILL Secretary-Treasurer

Canadian Graduate and Professional Student Survey (CGPSS) 2016

Geometric Predicates P r og r a m s need t o t es t r ela t ive p os it ions of p oint s b a s ed on t heir coor d ina t es. S im p le exa m p les ( i

PERMACULTURA INTENSIVE EDE. 3weeks,intensivetraining

3 2 - x is correct. Since 32 is the total games played, and x is her number of wins, the losses must be 32 take away x.

k g e a. m eri v al e 2 6 T H N O V E M B E R 2 N D D E C E M B E R M O R E G R E A T D E A L S I N SI D E

FOR SALE T H S T E., P R I N C E AL BER T SK

, _ _. = - . _ 314 TH COMPOSITE I G..., 3 RD BOM6ARDMENT GROUP ( L 5 TH AIR FORCE THIS PAGE DECLASSIFIED IAW EO z g ; ' ' Y ' ` ' ; t= `= o

Chemical Hazards and Hazard Communication

Recurrent Neural Network

Boy Scout Troop 41 Bay Village, Ohio A C K N O W L E D G E M E N T S. F i f t i e t h A n n i v e r s a r y C e l e b r a t i o n

Australia November 13, 2017

CONTRIBUTES TO LEED OBJECTIVES L

dependent and wr i t er i ndependent on-l i ne cur s i ve handwr i t i ng r ecogni t i on. Thi s

M M 3. F orc e th e insid e netw ork or p rivate netw ork traffic th rough th e G RE tunnel using i p r ou t e c ommand, fol l ow ed b y th e internal

Ash Wednesday. First Introit thing. * Dómi- nos. di- di- nos, tú- ré- spi- Ps. ne. Dó- mi- Sál- vum. intra-vé-runt. Gló- ri-

I n t e r n a t i o n a l E l e c t r o n i c J o u r n a l o f E l e m e n t a r y E.7 d u, c ai ts is ou n e, 1 V3 1o-2 l6, I n t h i s a r t

THIS PAGE DECLASSIFIED IAW EO 12958

AGREEMENT AND PLAN OF MERGER (Buye r Oriente d)

ATHLETI C SUPPLI ES & EQUI PMENT ( Li ne I t em) BID ID NO MI NI MUM ORDER REQUI REMENT OF $ BEGI NS: June 1, 2012 ENDS: May 31, 2013

Detect i on of Bra i n Damage in Psych i at r i c Popu l at i. Anton i o E. Puente. Depar tment of Psycho l ogy

In t e r n at ional Char it abl e Pl anning (wit h For ms )

T ensor N et works. I ztok Pizorn Frank Verstraete. University of Vienna M ichigan Quantum Summer School

Student Name: Date: Teacher Name: Micah Shue. Score:

I cu n y li in Wal wi m hu n Mik an t o da t Bri an Si n. We al ha a c o k do na Di g.

Wint er 20 18?Special Edit ion? Elect ion Guide

Welcome to the Public Meeting Red Bluff Road from Kirby Boulevard to State Highway 146 Harris County, Texas CSJ No.: December 15, 2016

WELCOME. O ne Vi si on Photography i s a award wi nni ng wed d i ng photographer & wed d i ng vi d eography i n S outh Wal e s

Snork Synthesis Lab Lab Directions

Transcription:

AI ML The Fut ur e of Assessment How wi l l AI, Aut omat i on, and Mac hi ne Lear ni ng Change How We Devel op and Del i ver Assessment s? Nat han Thomps on, PhD I nt er nat i onal Conf er ence on Educat i onal Measur ement, Eval uat i on, and Assessment

AI & Aut omat i on A 2016 r epor t f r om Del oi t t e est i mat ed t hat 40% of l egal wor k wi l l soon be r epl aced by AI I t i s not a quest i on of whet her aut omat i on wi l l happen i n t he assessment i ndust r y, but wher e?

What will AI al l ow us t o aut omat e? We l l be abl e t o aut omat e ever yt hi ng t hat we can descr i be. - St ephen Wol f r am...well, there is a l ot about t est devel opment and psychomet r i cs t hat we can descr i be!

Def i ni t i on AI : us i ng comput er s t o do t hi ngs t hat nor mal l y woul d expect a human ML : devel opi ng mat hemat i cal model s to trai n systems Aut omat i on: Usi ng computers t o r epl ace humans or make t hem mor e ef f i ci ent, but of t en j us t r ul e- bas ed

El evat e your pr of essi on Item writers Teacher s Essay mar ker s Test devel oper s Pr ogr am manager s Psychomet r i ci ans

Make a car not a horsel ess car r i age!

Make a car not a horsel ess carri age!

About ASC: AI & Mac hi ne Lear ni ng Founded 1979 out of t he Psychomet r i cs pr ogr am at t he Uni ver si t y of Mi nnesot a Sof t war e t hat makes i t easi er f or any org t o devel op qual i t y assessment s We ar e wor ki ng di l i gent l y t o expl or e t he oppor t uni t i es pr esent ed by AI / ML!!

I TEM BANKI NG: LEARN FROM YOUR BANK Pi l ot st udy on usi ng machi ne l ear ni ng t o anal yze i t em banks f ound we can pr edi ct 56% of i t em qual i t y j ust by anal yzi ng whi ch wor ds ar e used i n st em Words on difficult items Words on easy items Term t Term t been 3.886 happy -2.980 that 3.841 free -2.807 their 2.692 going -2.720 mother 2.669 bus -2.363

Cr eat e Mat r i x Prune Matrix Tr ai n Model 1. Pull item texts from FastTest 1. R e mo v e mi s s i n g d a t a 1. Recode as UTF 1. Cr eat e Document - T e r m Mat r i x whi c h maps wh a t wo r ds a r e u s e d o n whi c h i t ems 1. Remove al l wor ds f r o m ma t r i x u s e d l es s t han 5 t i mes 1. Reduces t he mat r i x f r om t hous ands of wor ds t o hundr eds 1. Cr eat e a machi ne l ear ni ng model t o pr edi ct i t em difficulty or item qual i t y based on f r equent l y used t er ms 1. Can t i p of f aut hor s on wor ds t hat make i t ems t oo eas y or har d Machine learning on item banks; system will learn what makes a good item and provide feedback to authors

I TEM BANKI NG Tr ack your wor kf l ow t o l ear n of bot t l enecks and best per f or mer s!

I t em Aut hor i ng: GUIDELINES St r ai ght f or war d but still under ut i l i zed: Gui de aut hor s on i t em wr i t i ng r ul es I mpl ement s i t em r evi ew at t he t i me of wr i t i ng

Item Banki ng: REI NFORCED L E ARNI NG What i f we asked t eacher s t o gi ve i t ems & t est s a t humbs up/ down?

AUTOMATED I TEM GENERATI ON APPROACH 1 APPROACH 2 St af f def i nes i t em skel et ons wi t h sever al var i abl es ( A year ol d man pr esent s wi t h chest pai n af t er... ) 6x ef f i ci ency Feed a t ext book t o an AI al gor i t hm and i t pr ovi des dr af t i t ems back t o you St i l l i n i t s i nf ancy but super exci t i ng SEE IN ACTION

AUTOMATED TEST ASSEMBLY Test Templ at es: f or ce al l t est s t o f ol l ow bl uepr i nt s t o ensur e cont ent validity Aut omat ed t est assembl y: Bui l d test forms di r ect l y t o bl uepr i nt s at cl i ck of a but t on

AUTOMATED TEST ASSEMBLY Cur r ent t ool avai l abl e f or downl oad Exper t s can use Li near Pr ogr ammi ng packages

TEST PUBLI SHI NG Eval uat e pool of i t ems t o aut omat i cal l y det er mi ne appr opr i at e CAT al gor i t hms and publ i s h a def ensi bl e CAT You s houl d not have t o wr i t e c ode t o publ i s h a CAT! Or do IRT!!!

MORE TEST PUBLI SHI NG St andar d set t i ng: Aut omat e how Angof f r at i ngs ar e gat her ed and t hen col l at ed i nt o a r epor t

MORE TEST PUBLI SHI NG Last year s wi nner of t he I nnovat i on Lab at t he ATP conf er ence: AI t hat wat ches how st udent s sol ve t ech- enhanced i t ems and t hen uses i t f or f ut ur e scor i ng. Si mi l ar t o AES. met acog. com

TEST DELI VERY Li near on t he f l y t est i ng ( LOFT) : Al l exami nees get 50 i t ems t o same bl uepr i nt s, but di f f er ent i t ems, f or bet t er secur i t y Comput er i zed adapt i ve t est i ng ( CAT) : The t est per sonal i zes i t sel f t o ever y exami nee. Per f ect mat ch f or per sonal i sed l ear ni ng. Mu l t i - St age Test i ng: Mel ds CAT, LOFT, and l i near t est s - per f ect f or l anguage assessment These ar e AI and aut omat i on based on mac hi ne l ear ni ng pr i nc i pl es ( I RT)!

REPORTI NG AND ANAL YT I CS Psychomet r i c Analytics: test and i t e m p e r f o r ma n c e r epor t s r epl ace a ps yc homet r i c i an wi t h an al gor i t hm

REPORTI NG AND ANAL YT I CS Au t o ma t i c a l l y mo v e psychometri c stati sti cs bac k t o i t em banker Tr ack al l hi st or i cal stats to fl ag possi bl e dr i f t

REPORTI NG AND ANALYTI CS: L OCAT I ONS / P R OGR AMS Ca n we c onnec t a busi ness i nt el l i gence engi ne t o make i t easy f or st akehol der s t o do t hi s however t hey l i ke?

REPORTI NG AND ANALYTI CS: Psychomet r i c For ens i c s Perform psychomet r i c f or ens i cs t o spotl i ght i ssues of test security and val i di t y threats Can we do t hi s i n r eal t i me?

AI & ESSAY I TEMS Aut omat ed essay scor i ng - pr ovi des a f r ee second or t hi r d r at er t hat s been shown t o be as accur at e as humans Two appr oaches: base i t on l anguage t heor y, or on pur e dat a sci ence And yes, t hey exi st f or Ar abi c! Al so: what can t he FACETS appr oach t el l you about r at er s, pr ompt s, and r ubr i cs?

AI & STUDENT FEEDBACK Cogni t i ve di agnost i c model s: pr ovi de det ai l ed f eedback on i ndi vi dual ski l l s bei ng t est ed Theor y- dr i ven machi ne l ear ni ng appr oach

MORE! Thi s i sn t even t ouchi ng on AI / ML i n t he use of t est scor es Adapt i ve l ear ni ng Recommender syst ems Predicting job performance and other workforce out comes What can per sonal i t y aspect s pr edi ct about l oan def aul t s? Rent er saf et y?

THANK YOU nthompson@assess.com