Designing Rubrics For Classroom Assessment Professor Timothy Farnsworth, CUNY Hunter College Communicative Assessment Is Hard! We are making inferences about what people can and can’t get done in the real world, not: What grammar or vocabulary they can choose from 4 options on a MC test Tasks must be realistic but can’t be TOO realistic: 2 Time constraints Stakes are different Interlocutors are different Setting, format usually different Speaking Assessment Challenges Speaking is the hardest skill to assess Impermanent -> must be rated on the spot or recorded Interactive -> Real communication takes two persons Nonverbal -> Much real communication takes place without language use Topical -> Students’ knowledge base affects their speaking output 3 Oral Assessment / Testing Tasks Must get students involved in speaking! In communication, many things happening Listening Nonverbal channeling Decisions made about what to say, what not to say Students practice avoidance strategies, game the system, don’t take risks Assessment tasks take many forms 4 Types of Assessment Tasks Many options – some common ones: Interview, either scripted or unscripted Picture description Role Play Pair or group tasks Almost any classroom task, really! 5 Sample task: Pair oral assessment Each student has a different picture; communicate to find similarities and differences 6 Scoring tasks and communicating results We need some way to turn the “observation” into a “score” or “grade” Usually done via numbers, can be done as letters, smiley faces, etc. – doesn’t matter Should be clear, simple Should be “reliable” -> repeatable, dependable Should be “valid” -> meaningful Shouldn’t cause any harm 7 Types of scoring rubric Holistic Analytic Task Fulfillment (“primary trait”) Performance decision trees Checklists 8 Score Description Communication almost always effective: task performed very competently Holistic Rubrics Functions performed clearly and effectively 60 response to audience/situation OneAppropriate overall score tied to a “performance descriptor” Coherent, with effective use of cohesive devices of linguistic features MostUsefamiliar type of almost rubricalways effective; communication not affected by minor errors Assign score of the description that best matches ….. performance Communication somewhat effective: task performed somewhat competently Problem: Performance doesn’t match description at the Functions performed somewhat clearly and effectively same level inappropriate all areas Somewhat response to audience/situation Somewhat coherent, with some use of cohesive devices Advantage: Simple to somewhat use andeffective; understand Use of linguistic features communication sometimes affected 50 40 by errors 30 … 20 No effective communication: no evidence of ability to perform task No evidence that functions were performed … 9 Analytic Rubrics Level 0 1 Accuracy Fluency Pronunciation Two moreresponse; categories of scores Noor structured Speech is halting, Wholly or partly unintelligible single words only long pauses due to pron. Errors Usually based around linguistic categories such as pronunciation, grammar, etc. Also quite common Multiple inaccuracies in Pauses occur but Pron. Errors increase difficulty vocabulary choice, are not overly long, for hearers Assign scores for eachspeaker category per performance sentence structure does not inordinately ‘word Problem: Can be cognitively challenging for raters search’ Advantage: provides more detailed information 2 Few major errors wrt vocab or SS 10 Normal or nearnormal speed and flow of speech Pronunciation is native-like or does not increase difficulty for listeners Task Fulfillment Rubrics Score UsuallyDescription look like holistic rubrics 4 InsteadCandidate was polite, friendly, guest successfully, of describing language, theyhelped describe whether the to the complaints studentresponded completed “task” appropriately, asked for clarification, etc. 3 2 1 0 Usually used in role play type assessments Useful … in testing for employment, other settings Useful … when there is a specific set of tasks that involve language and canappeared be specified Candidate rude, could not help guest Example rubric fordid hotel employee successfully, not respond to testing: complaints appropriately, could ask to for respond clarification, etc. Job applicant is tested fornot ability to guest requests and complaints, answer telephone, etc. in role play situation, No response 11 Performance Decision Trees Fairly new to the scene (Fulcher 2011) Use a flowchart design Good for clear, simple scoring when the task is very consistent / clear Give points for each successful “yes” on the flowchart Lets the test raters answer very specific questions 12 13 Self / Peer Assessment Students have some good ideas about their and their classmate’s performances In a busy classroom, teacher cannot hear everyone Students can use any feedback they get Training students to listen carefully and help each other is good practice in general But: Students are not always good evaluators of their peers May not catch pronunciation or other important mistakes May be LESS proficient than the person they are supposed to assess Only ask them to assess in areas they can definitely “do” – yes / no checklists are great for this 14 Checklist Rubrics Phrased very simply like a can-do statement: “Student can understand and use vocabulary for classroom objects.Yes / No / Partially “Student can use formal language in giving oral presentations” Yes / No / Partially Can be broken into parts: “Student understands WH questions” Yes / No / Partially “Student responds to WH questions using appropriate grammar and vocabulary” Yes / No / Partially “Student responds to WH questions at length and returns questions appropriately” Yes / No / Partially 15 Checklists for student self-evaluation Self Evaluation – teacher collects evidence Checklists are simple and clear (example) List main objectives for the week, create checklist for students Have them do and turn in anonymously Use the info to plan review sesssions in future 16 Keys to consistent, fair scoring 1: KISS principle for tasks Keep It Simple, Sweetheart! Don’t use any more complicated assessment than you really need One or two tasks is enough as long as you are getting the right kind of interaction Make sure you get “enough speech” to make a rating using the right rubric Make a system that everyone can easily understand Good speaking testing usually can be done in less than five or ten minutes More “high stakes” testing takes longer, up to an hour 17 Keys to consistent, fair scoring 2: Use the simplest rubric that works I like checklists or performance trees for classroom and self assessments I like analytic rubrics for more high-stakes testing They give information on what things the student needs to work on Holistic rubrics don’t provide much information for students I don’t like task fulfillment rubrics for most situations 18 It’s not clear from the score what the student needs to work on It’s sometimes possible to do a task using little language; the task fulfillment setup still allows high scores Keys to consistent, fair scoring 3: Write the rubric so students can understand it Use plain and simple language Use “I can / can’t do X” or “You did X” when possible Explain the scores and the rubric to them before the test – no reason for it to be a secret They should understand why they got their score and what they can work on next time Do not use a self assessment for a grade Do not use peer assessment for a grade 19 Your assignment: Write a scoring rubric Find a partner next to you Together, pick one rubric style: Holistic, analytic, decision tree, checklist, task fulfillment Think about students in a typical class you teach: 20 What level are they? What do you need to learn more about in terms of their speaking skills? Write a rubric with a few levels and clear descriptions Compare your rubric with a pair near you – does it look similar? Why or why not? Sample task: Pair oral assessment Each student has a different picture; communicate to find similarities and differences 21