AY 375 - Fall 2014: Sixth Day Lesson Plan

Today we'll be discussing how to write good multiple choice and free response questions. We will also discuss grading rubrics and grading.

1. Deeper conceptual knowledge CAN be probed with multiple choice questions, but writing effective questions takes time and practice.
2. Especially when writing free response questions, it can be useful to develop a grading rubric for each question as a way of ensuring that your questions are specific, clear, and not testing the same concept over and over again.

First have students discuss in pairs how section went (5 min).

Open the floor up for general questions and sharing about how sections are going. Some questions include:

• What did you do?
• How did you implement your activities?
• What worked?
• What didn't work?
• What would you do differently?
• How did you assess learning?
• Did you receive any unexpected questions/reactions/etc.?
• Did anything unexpected happen?
• What were you thinking about while you were running section? Any moments of panic?

The students should have prepared one multiple choice question and one free response question as last week's homework.

1. (10 min) Share your questions with a partner. Have them attempt to answer the questions or at least determine what concepts were being probed and what would have constituted as an “acceptable answer.” Once you have gone over all the questions, discuss in your small groups the goods and suggested improvements for the questions. Some questions to consider (perhaps project on the board):
   1. For both:
      • Is the wording clear?
      • What concepts are being tested?
   2. For the multiple choice:
      • Is there clearly only one correct answer?
      • Are there any obvious throw-away answers?
      • Are you able to rule out any response because of the wording alone (i.e., are there hints in the structure)?
   3. For the free response:
      • Does part B test the small conceptual/procedural knowledge as part A?
      • What if students cannot solve part A? What does that imply for part B?
      • What sort of responses might students give under the pressures of an exam setting?
2. (5-10 min) Come back as a class and discuss.
   • Did you learn anything surprising?
   • Is this easy? (Unfortunately, NO!)
   • What part of question writing did you find the most difficult?
3. (20-25 min) Go through question example slides as a class exercise.

Some notes on multiple choice questions:

Despite their outward appearance, these questions are actually inherently nonobjective. Grading an essay exam is subjective to the personal feelings of the grader, compared to running a Scantron through a machine. Grading written problems falls somewhere between the two. This is only partially correct, “grading” a Scantron is completely objective, but the subjective aspect of multiple choice questions comes in the creation of the item (the question), the response (the correct answer), and the distractors (the incorrect choices). If everyone in the class was to write a question about the Doppler Effect, we would see a range of different questions and a range of ideas probed. That is subjective.

The ultimate goal of testing is to measure what the students actually understand, and the process of interpreting the meaning of a student's response to a MC test is a subjective one. There are three major issues behind writing these sort of questions:

1. the physical format and layout of the question
2. the conceptual hierarchy of the questions
3. the statistical item analysis

Below are some guidelines for each of these items.

You are testing the students' understanding, not their reading ability. Long passages of text cause slow readers to skim and often miss details. Questions should not include strings of prepositions, parenthetical statements, or extended clarifications. Misinterpretation is impossible to completely predict, but concise, clear questions can do a lot to minimize the chance of students misreading the question or the response. For example,

You forget that the star Betelgeuse is a red giant and apply the method of magnitudes to determine its distance. The true distance to Betelgeuse is actually… (a) shorter than you calculated, (b) the same distance that you previously calculated, © farther than you calculated.

Is short, to the point, and clear. You might be tempted to elaborate on small points that are not the main conceptual item that is being tested, but care must be taken. For example,

You forget that the star Betelgeuse is a red giant (a very luminous star in the top right of the HR diagram with relatively low surface temperature) and apply the method of spectroscopic parallax—a comparison of the star's apparent magnitude, estimated from the HR diagram, and its absolute magnitude—to determine its distance from Earth, which can be considered the sam as its distance to the Sun because the Earth-Sun distance is negligible given the scales involved. The true distance from Earth to Betelgeuse is actually…

In an attempt to be completely clear, the stem has become more difficult for most students to understand!

Over the years, students have learned that when novice faculty includes choices such as “all of the above”, these are frequently the correct answer. It's easier as a test writer to write correct statements than to come up with plausible sounding incorrect statements. Also, students have also learned that the longer answers are usually the correct answers. You can avoid these situations by making sure your choices are all of similar length, contain a similar amount of scientific vocabulary, and ensure that an equal number of choices (A), (B), etc. are correct on the overall test.

Consider

The thermo nuclear reactions in a stellar core are the result of (A) fission, (B) fusion.

This tests what? Unfortunately, it tests word association. A student can get by without knowing what fusion means. Students learn to adopt a strategy of memorizing definitions and words, rather than understanding concepts. As another example,

The Monotillation of Traxoline (attributed to Judy Lanier) It is very important that you learn about traxoline. Traxoline is a new form of zionter. It is monotilled in Ceristanna. The Ceristannians gristerlate large amounts of fevon and then bracter it to quasel traxoline. Traxoline may well be one of our most lukized snezlaus in the future because of our zionter lescelidge.

Directions: Answer the following questions in complete sentences. Be sure to use your best handwriting. (1.) What is traxoline? (2.) Where is traxoline monotilled? (3.) How is traxoline quaselled? (4.) Why is it important to know about traxoline?

Notice how easy it is to get 100% without understanding a single thing about the passage! Pay attention to wording.

Less applicable for GSIs, but if particular questions are frequently missed, it is necessary to probe whether they are missed because of a lack of clarity, or because the question is testing difficult concepts. If the former, revise!

See below for some general bullet points on what to look out for.

Some notes on free response questions:

These notes are written with “Astro C10 quizzes” in mind, but the ideas are generally applicable.

• These points are valid for both quizzes and exams:
  • Test the material emphasized - Exams should reflect the fact that students should know the big concepts really well, as opposed to knowing a bunch of smaller concepts only peripherally.
  • Keep questions short and to the point - Students should spend the majority of their test time thinking and answering/writing, NOT reading.
  • Edit questions for clarity - Clear questions tend to be shorter and if anything is ambiguous it confuses and slows down students and makes it harder for you to grade it fairly. Have someone else take your exam to give some feedback. If you're taking an exam for someone else, be critical and think about possible ambiguities.
  • Don't write a long test - Keep it concise, to the point, and clear! The rule of thumb is your average student will take double or triple the time it takes a GSI to complete the exam. Also, 90% of your students should finish the exam completely in the allotted time.
• Quizzes vs. Homeworks
  • They're quite similar in their construction and type of questions.
  • The main difference is that quizzes should be shorter and have easier questions, since students have much less time to work on quizzes and must work on them alone.
• Quizzes are:
  • short
  • usually given in section
  • questions are of exam difficulty (i.e. easier than homework questions)
  • Not too in-depth or calculation-based. Some light calculation might be OK, but take care. (Not everyone will remember a calculator no matter how many times you remind them, and many people will be petrified of the idea of computing things on their own.)
  • usually only cover recent material (i.e. the past 2-3 weeks)
• Quizzes are used:
  • by the GSI to gauge each students' understanding of recent material individually (as opposed to homeworks which can be done in groups)
  • by the student to gauge their own understanding of recent material and get a feel for what a college level intro science course non-Scantron exam will look like and what level of understanding they are expected to have for the exams
  • by the prof to get a grade early in the semester that's more important than a single homework, but isn't the big production that an exam is (you might not have covered enough material for a full exam)
• What makes good ones?
  • not too long (both in length of individual questions and number of questions)
  • not too hard
  • not too easy (shoot for a variety of difficulties in questions)
  • relevant to recent material
  • varied in the types of questions (multiple choice, fill in the blank, calculation/mathematical, read a graph, free/paragraph response)
  • unambiguous with easy-to-read questions
  • not mathematically demanding - questions should probably not require a calculator and should definitely not include extensive tedious calculations
  • representative of same knowledge required for exams
  • gradable for partial credit (not simply binary right/wrong like Scantron exams)
  • specific about what you're looking for in free response type questions: Don't give students the opportunity to 'core-dump' for a problem, it wastes their time spewing forth useless information and makes your life tougher when you have to grade the mess.
  • quick to grade (this makes your life much easier and helps the grading be more fair for all of your students)
• Quizzes are meant to be relatively low stress (especially compared to full exams)

General notes from previous years in convenient bullet-point form:

• Exams exist to:
  • Evaluate student learning for University-required grading
  • Motivate students to study and understand the material
  • Allow the instructor to evaluate his/her progress educating students about the material
  • Provide feedback to students about their understanding and study habits and illustrate specific gaps in their understanding of the material

• What makes good multiple choice exam questions?
  • Not too long/wordy (neither the question nor the possible answers)
  • Not too much calculation
  • Not too tricky (i.e. there shouldn't be two extremely similar answers)
  • Relevant to important material (as opposed to really obscure/minute details)
  • Very clearly written, precise wording in both question and answers
    • E.g., 'Which best describes…' as opposed to 'What is…' or 'How does…happen'
  • Pedagogical as well as evaluative (e.g., some questions should probe common misconceptions)
  • Questions that are very easy or very difficult are OK as long as the test has questions with a variety of difficulties

• What makes bad ones?
  • Long answers!
  • Excessive use of 'all of the above' (some people say any use of 'all/none of the above' is a bad thing)
  • Questions that can be solved without knowledge of the material (usually because of the use of too many blatantly wrong or “funny” possible answers)
  • Multiple potentially correct answers (usually from vague questions or possible answers)

• Non-multiple choice questions
  • Types of questions:
    • Matching
    • Fill-in-the-table/blank
    • Simple calculations
    • Short answers and paragraph/free responses
    • Diagrams, plots, graphs
  • Most of the same points discussed above apply here:
    • Questions should be clear, easy to read, and unambiguous
    • Questions should be relevant to the material presented and emphasized (do not test on obscure passages of the textbook)
    • For high-value questions, allow for partial credit
    • Make the questions easy to grade! Don't give students the opportunity to 'core-dump' for a problem: be very specific about what you're looking for in these questions.
  • Can be more time-consuming than MC questions, depending on the overall length of the exam.

• Activity in groups of 3. Each group receives the same free response question.
• Have each group develop a key and rubric for the question. (5 minutes)
• Give each group one student response to that question (three different responses, one for each group). Have each student grade the quiz based on that rubric individually. (5 minutes)
• Have students compare your grade with others in the group and discuss. (10 minutes)
• Discuss as a class and recap the main ideas of grading as a class. (10 min)
  • emphasize that reading some responses first is important.
  • discuss how to handle erroneous info.
  • The most important part of grading: Grade fairly and consistently for ALL students.
  • Try not to look at student names while grading anything.
  • Grade in blue or green. My (Aaron's) philosophy: Grade your own work in red, because you don't care about your own feelings. Grade other people's work in blue or green.
  • GRADE WHAT YOU TEACH AND TEACH WHAT YOU GRADE.

• Most questions should have 1 and only 1 correct answer (matching, fill in the blank, put in order, etc.).
• Paragraph or few sentence responses or plotting can be uglier.
• Try to give partial credit where you can. Always give points for correct steps even if the final answer's wrong. If they get the final answer but their steps or logic to get there is wrong, give them some points, but not too many.
• Obviously if they screw up part (a) by a factor of 2, but carry that extra factor through parts (b) through (f) and get everything else right (while including the factor of 2), they should only lose points on part (a). Also, stress this fact to your students so they don't get frustrated if they can't do (a), but the rest are doable (maybe even tell them to make up an answer to use for later parts, or in the question say 'use 5km for the rest of this question if you don't get part (a)').
• In longer answers, you should usually reward for correct information more than you punish for incorrect information. With that said, if they say something really wrong or even contradictory to the rest of their answer, they should be penalized a decent amount.
• Hopefully on your quizzes and exams you stress to students that they must write clearly and explain their steps and logic clearly. If you can't read their writing or understand what's going on, you should usually assume it's wrong.
• Be suspicious: If you see similar, very wrong answers, flag the tests and compare their answers to other questions. Hopefully you can look out for cheating while the quiz/exam is actually going on, but you won't be able to see everything.
• Talk (probably through e-mail) to students in your section(s) who performed very poorly (grades of less than 40% or 50%). They may be too shy to ask for help even if they know they need it!

• You'll grade these kinds of tests with your fellow GSIs.
• Have a few people people double-check the answer key Scantron for each version of the exam BEFORE you start running all the tests through the machine.
• Note any questions that are missed quite frequently. Reasons for this can include:
  • The question was poorly written or possibly too hard
  • The students just didn't know that material very well
  • There's an error on the answer key
  • Something went wrong with the machine or the answer key Scantron
• Note any one student who missed a TON.. Reasons:
  • They really just don't know what's going on
  • They marked the wrong test version (you might be able to re-run it through the machine with the correct answer key or their GSI or the Head GSI might have to grade it by hand)
  • They used some writing utensil that the machine doesn't like (their GSI or the Head GSI might have to grade it by hand)
  • Their test is too wrinkled or has coffee spilled on it or whatever (their GSI or the Head GSI might have to grade it by hand)
  • Use your judgment here; it's annoying to grade a Scantron by hand, but if they really just accidentally spilled something on it, then you should probably grade it by hand with no penalty. If they're dumb enough to mark the wrong test version, well, they might deserve some extra points off for that.
• It's a good idea to skim over each of your student's tests to see if there were any obvious bad erasure marks or anything like that which may have led to an answer being marked wrong unfairly. However, don't feel bad if you don't catch every one of these – your students will not miss any!!
• Like after quizzes, strongly consider talking to students in your section(s) who performed very poorly (grades of less than 40% or 50%).

• Usually not the GSI's responsibility to grade.
• Basically, all of the above rules apply.
• Cheating is certainly more of an issue since you can't watch everyone do their homework, so beware! As mentioned before, hopefully the graders will flag any possible cases of cheating and then you can take a closer look at the actual assignments and decide whether or not any academic dishonesty actually occurred.

• You should come up with some objective grade calculation (unless one is already provided to you by the prof and/or Head GSI).
• A grade for Discussion Section might include (ONLY if these aren't assigned grades on their own elsewhere in the course grading rubric):
  1. attendance in section
  2. participation in section (group work, coming up in front of the class, asking the GSI questions, worksheet diligence, etc.)
  3. in section quiz(zes)
  4. star party attendance
• Usually this isn't a huge part of the overall course grade, but it's sometimes a non-zero amount of points and you should be fair and consistent about assigning them.
• With that said, if someone really went above and beyond and worked really well in section, usually you have the authority (as long as it's cool with the prof and/or Head GSI) to give them a little something extra in their section grade (though you probably shouldn't give them over 100% of the section points).
• On the other end of things, sometimes students with borderline grades will try to wring a few extra section points out of you to push their grades up to the next letter.

• Usually GSIs don't have much control over this. However, GSIs occasionally help decide the course grading rubric before the semester starts.
• The course grading rubric should be well-defined and a (relatively) simple calculation.
• It should be spelled out in detail on the syllabus (which should be handed out or posted online at the very beginning of the semester).
• Converting numbers to actual letters is the hardest (and usually the most mysterious and opaque to students) part of this process.
• Finally, most profs/Head GSIs will let each GSI have a small amount of discretion in final letter grades (with pluses and minuses) for their students who are right near a letter grade cutoff.
  • Most profs/Head GSIs will try to choose the cutoffs such that no one is close to any of the boundaries.
  • However, this discretion is fairly common practice since the GSI should know the student (and their performance in the course) better than either the prof or Head GSI.
  • Thus the GSI can use the student's performance in their section to decide whether or not to bump them up above or below the cutoff.
  • With that said, unless the student was absolutely horrific in section, GSIs should probably never bump anyone down below a cutoff!

1. Draft a one-page teaching philosophy.
2. Continue to improve your two exam questions. Email me one of your two for further feedback.
3. In the next few weeks, try a new group work strategy or method of discussion in your sections or office hours. You'll report back to the class on how things went in a few weeks.