IImproving Students’ Lab Practices: the Performance Grade
G.L. Lippi , Universit´e de Nice Sophia Antipolis,Institut Non Lin´eaire de Nice CNRS, UMR 73351361 Route des Lucioles, F-06560 Valbonne, France ∗ (Dated: November 9, 2018)Instilling good laboratory working attitudes in students is a difficult but very important task,especially in the first level courses. The introduction of a grade, based on the observation of workpractices during laboratory sessions, can be strongly beneficial towards the acquisition of positiveskills covering not only the technical aspects, but also the acquisition of both independence and teamwork. Explicit suggestions are given for basing the grade on specific observations and a quantitativeanalysis is performed to guarantee that the higher intrinsic volatility of the Performance Grade doesnot affect the final laboratory grade. I. INTRODUCTION
Bad habits are better avoided than unlearnt: this dic-tum applies to laboratory practices as well. Instructorsare often confronted with the problem of how to con-vince and/or encourage students to take a positive at-titude towards their practical experience and maximizetheir learning, rather than concentrating their attentionon the short-term reward: the grade.Given that laboratory work is for the most part a rel-atively new and foreign experience for a large numberof students beginning their studies in Physics (and moregenerally, Science), on the one hand one has the privilegeof building upon a (fairly clean) slate but, on the otherhand, one also encounters the difficult task of convincingthe students to take up good habits. Although buildingon rather pristine ground may sound like an ideal situ-ation, this often turns out not to be the case, since theaverage student does not have a good idea of what ex-perimental work truly means. Thus, at times it may berather difficult to make oneself understood. Compoundedwith the pressure of obtaining good scores, the commu-nication hurdle can become frustrating for the instructorand damaging for the student.For all the explanations about what is expected, inan average class the instructor finds herself/himself con-fronted with the (hopefully only) occasional case of thestudent who at the end of the course still has not got itright. And this is not necessarily for lack of trying fromeither side. Indeed, the feedback to the student arrivesat best at the following session (typically one week later)and since laboratory courses – in particular at first level– tend to have a large attendance, the instructors maychange from one session to the next. Thus, the transmis-sion of information becomes problematic and it is all toonatural that some students may take a “wait and see”attitude.The standardized system of education, as it is univer-sally employed at least in western countries, uses a grad-ing system not only as a form of evaluation, but also offeedback. In spite of all the possible criticism which canbe raised to its use, it is certain that students are well attuned to the grade, since they have beeen exposed toit through all their schooling. Thus, the idea of using aform of grading to give feedback to students and signalwhether they are on the right track seems to be a rea-sonable one. After all, this is what is done in theoreticalcourses where homework is graded weekly.The practical objection which is often raised is how todo this in a fair and reliable way. I am going to present aframework in which such a grade, which I am going to callthe performance or participation grade , can be satisfac-torily used for the benefit of students without impingingon the fairness of the overall laboratory course evaluation.I have introduced this method of encouraging students’participation and guide their laboratory work about tenyears ago in undergraduate labs in our Physics programand have worked with numerous collegues – among oth-ers several Teaching Assistants and Temporary Instruc-tors. The experience has been positive and has allowedto more effectively guide a much larger percentage of stu-dents towards developing a positive working attitude inExperimental Physics. II. THE PERFORMANCE OR PARTICIPATIONGRADE
The Participation Grade is an evaluation given at theend of each laboratory session by the Laboratory Instruc-tor(s) of the quality of the work performed by each in-dividual student in the course of the session. Its mainfeatures are summarized in Table I. The main goals to beassessed are the student’s attitude towards learning andher/his active involvement in experimental work whichshould progressively lead to the acquisition of the auto-matic reflex of continuously questioning both proceduresand results, rather than passively accepting any obtainedresults or all the instructions received. This in turn de-velops critical thinking and builds a strong foundation forthe development of the basic skill expected of any physicsgraduate: the ability to analyze new situations and findsolutions to new problems. Of course, the degree of suc-cess has to be reasonably assessed in accordance with a r X i v : . [ phy s i c s . e d - ph ] M a r TABLE I: Synoptic definition of the characteristics of the
Performance or Participation Grade . The first category (topbox) highlights the goals each student has to strive for, thesecond box the nature of the grade (individual, even thoughthe work is collective), the last three categories present thestrenghts and weakness of this kind of grade.AttitudeQuestioningGoals Critical thinkingInteraction and group workConsulting sources (advanced labs) Nature Individual gradeEffectiveness Quantitative assessmentImmediacy Feedback at the end of sessionWeakness Subjectivity and reliability (fluctuations) the course level (i.e., introductory vs. advanced lab) butclearly requiring this from students is beneficial at all lev-els. Another important point in the evaluation is the kindof interaction that each individual student has both withthe Instructor and with her/his Peers: a student alwaysasking for help, for the right answer , for what needs tobe done now is clearly on the wrong track. In the sameway, a student passively following a leader – e.g. thelab partner – is not standing up to the expected level.However, as stressed in Section V, a student leading thegroup by neglecting her/his partner(s) is also not per-forming up to standards, since Laboratory work is sup-posed to be a group effort. This setting offers therefore aunique opportunity to learn about (scientific) interactionand collaboration.By its nature, the Participation Grade is individual,even though laboratory work is of a collective nature.The reason for this apparent dichotomy is that throughcollective work each student has to conquer the outlinedgoals, thus the feedback provided by the Performancegrade must be tailored to the individual. Students maybe at first surprised to be individually evaluated for groupwork, but once the scope is clearly explained they willappreciate its importance and will react positively. Thisgrade operates therefore as a positive feedback mecha-nism for each individual within the ensemble. Transmit-ted to each student at the end of the session, togetherwith clear explanations and specific recommendations (inoral or written form), this assessment gives an immediatereturn on the less-tangible goals (i.e., those not related togetting the correct results from the experiment), by pro-viding individual and specific directions to acquire goodlaboratory practices.The Instructor’s assessment should be based on twobroad categories of indicators , summarized in Table II.The actions which are listed in the table are a non-exhaustive list of the points which can and should beexamined by the Instructor. The breadth of the task, al-though mitigated by the grading scheme suggested below TABLE II: Global (but not necessarily complete) summary ofInstructor actions suggested for assessing the student’s Per-formance Grade. The left column classifies the actions asindirect (observation without interaction with students) anddirect (resulting from direct interactions). Both kinds of ac-tions should be performed by every Instructor during each labsession for the students of which s/he has charge. The rightcolumn gives a non-exhaustive list of specific points to be ob-served for arriving to the grade assessment. The symbols “+”and “-” in front of various points in the analyical discussion(r.h.s.) stand for positive and negative attitudes which shouldeither be rewarded or sanctioned. As mentioned in the text,these are guidelines rather than rigid sets of checkpoints, andthe grade should be given in broad categories (thus mitigatingthe burden on the Instructor).Action DescriptionObserve the interactions among students:Indirect any member leading and/or dominating the group?any member left out?Observe the degree of involvement:is everyone participating in the work?Indirect to what degree? + taking turns+ discussing the work- just writing the results- sitting back and waitingDirect Kind of questions asked of the Instructor:+ trying to understand?+ actively looking for problem?- asking for a ready solution?- prying out the right answer?who is participating in the exchange?Direct Questions/challenges posed by the Instructor: a. globally, to the group b. directly, to an individual studentTechnical Instrument operationUncertainties and tolerancesPitfallsFundamental Understanding of processesComprehension Intuitive descriptionFurther developmentsProblems (Section III), immediately indicates the difficulty in giv-ing a quantitatively reliable grade. Fluctuations must beexpected, due to the sometimes large number of studentsand groups (and even different kinds of experiments asingle Instructor has to follow), to students’ behaviour(some groups will be more demanding than others), tomore or less frequent requests for help by some groups,to the need for troubleshooting problems with instrumen-tation, etc. Thus, the task appears to be daunting andmay discourage anybody from using this kind of gradingscheme, were it not for the fact that it is possible to setup a weighting system which renders it at the same timeeffective for the student and practically uninfluential forthe final grade, thus lifting any possible anxiety the In-structors may feel about using it. In the following we willprove how it is possible to set up the scheme in such away that the reliability of the course grade is not affectedby the use of the Performance Grade.Achieving this goal will also ensure that any criticismrelated to the higher intrinsic volatility of the Perfor-mance Grade, due to the perception-based assessmentgiven by the Instructor rather than on grading a writtendocument, will lose any value and the full benefits of thisgrade may be reaped. III. GRADE CONSTRUCTION
In the context of the grading scheme, and in orderto simplify the discussion, we are going to consider thetotal grade assigned to each student for a LaboratoryCourse to be composed of an Ordinary Grade, G o , and aParticipation (or Performance) Grade, G p : G t = M (cid:88) i =1 w o G o,i + N (cid:88) j =1 w p G p,j , (1)where w o and w p are the weights which are assigned tothe ordinary and participation grades, respectively. Forthe sake of generality, and in order to exploit one ofits advantages for improving the total grade reliability,we assume that the total number of ordinary grades be M (cid:54) = N , with N number of participation grades . Themeaningful inequality is M < N , which is equivalent tosaying that the lab is organized in such a way as to givemore than one participation grade per lab session (cf.discussion in Section IV).Throughout the paper we consider the total grade G t normalized to 1 (maximum value) and use percentages torepresent actual grades. This choice has the advantageof rendering the discussion independent of the gradingsystem, which changes from one country to another (andmay not even be uniform even among different Universi-ties of a same country). The Reader will make the nec-essary adjustments to convert the values given into theunits of her/his University for specific use.The simplification that we introduce in grouping under G o all the aspects of ordinary grading does not restrictthe generality of the discussion, but simply serves to setout G p from the rest. If the Ordinary Grade is com-posed, for instance, of graded Lab Reports and a finalexam, or of grades on the Labbook and separate LabReports, or whatever other combination, then it sufficesto decompose G o further (an example of more complexgrade composition has been discussed in the context ofgrading accuracy ).Fluctuations (even strong ones) in the PerformanceGrade have to be expected. It will be therefore conve-nient for the Lab Supervisor to give the team of Lab In- structors some common rules for the Participation Grade,specifying semi-quantitative criteria based on Table I (oron variations of the criteria exposed there). Predeter-mining a small number of possible grades not based onthe usual grading system but oriented towards more con-structive feedback for the student is also quite helpful inestablishing the Performance Grade. A possible set maybe: Excellent, Good, Acceptable and Insufficient . Thesebroad categories – which can be cast in different form(e.g., letters in a numerical grading system) – have thedouble advantage of: 1. partially diverting the students’attention from the detailed grade and 2. allowing, whenneeded, for grade renormalization when several Instruc-tors participate in the lab supervision. For instance, aLab Supervisor who notices that particular Instructorstend to give excessively high, or low, evaluations com-pared to their collegues, may weight the ParticipationGrades issued by the different collegues before transform-ing them into the usual grading scale.In the following section we discuss in detail the influ-ence of the (admittedly larger) fluctuations in G p on theaccuracy of the global grade G t and what constraints canbe applied to ensure that these fluctuations do not im-pinge on standard deviation, σ G t , of the grade, i.e., on itsreliability. One additional precaution is to plan the grad-ing scheme with the largest reasonable number of valuesfor G p , since in the averaging process one can expect toobtain an estimate for G p whose reliability grows withthe number of events. Changing Instructors, wheneverpossible, also ensures better averaging. Indeed, as dis-cussed in the following section, it may be useful to have,if possible, two Instructors assign – independently of eachother – a G p to each student for every lab session. IV. IMPACT OF THE PERFORMANCE GRADEON THE FINAL GRADE AND ON ITSACCURACY
An assessment of the impact of the participation gradeand its uncertainty on the reliability of the global gradecan be obtained using an a priori analysis . Using theproposed grade composition, eq. (1) with w o + w p = 1,we further assume that each grade be attributed withrespective uncertainty σ o and σ p . To simplify the discus-sion, we assume each uncertainty to be homogeneous overthe ensemble of grades of the same kind (e.g., all Perfor-mance Grades are affected by the same σ p ). However,this constraint does not restrict the validity our analysisand, if needed, can easily be relaxed . Notice that byusing the statistical description of the uncertainties, weimplicitely attribute a gaussian nature to the problem .Our discussion is based on the size of the uncertaintyon the participation grade σ p relative to σ o . These uncer-tainties can only be estimated a priori and in order tofix ideas , we will set 2 σ o = 0 .
03. This choice amounts tosaying that the ordinary grade G o is guaranteed within ±
3% with probability 95%, the usual error margin setin risk analysis (as discussed , Section 4) – nearly 100%(more precisely 99.7%) would require ± σ . The follow-ing analysis, however, is entirely general and holds forany value of σ o . This example is just a way of having areasonable reference for interpreting the discussion whichfollows.The a priori estimate for σ p in principle requires directknowledge of the conditions under which the supervisionis conducted. However, as we show below, it is possible todetermine conditions ensuring that the variability of theParticipation Grade does not reflect onto that of the totalgrade G t , whose uncertainty remains solely determinedby σ o .Following , we easily arrive to the expression for thecombined uncertainty on the global lab grade σ t = (cid:114) w o σ o M + w p σ p N , (2)= w o σ o √ M (cid:115) MN (cid:18) w p w o (cid:19) (cid:18) σ p σ o (cid:19) , (3) ≈ w o σ o √ M (cid:34) MN (cid:18) w p w o (cid:19) (cid:18) σ p σ o (cid:19) (cid:35) , (4)where in writing eq. (4) we have implicitely assumed therelative weight coefficient of the participation grade, w p w o ,to be small so as to use a first-order approximation tothe square root.In the present context, the approximation is not a sim-ple mathematical assumption to simplify the calculation:its nature reflects the requirement that the participationgrade should not impinge on the accuracy of the gradeand represents the constraint that we want to impose toensure fairness. Thus, we require the second term in thebracket, eq. (4), to be small. For the sake of generality,we fix its value to be12 MN (cid:18) w p w o (cid:19) (cid:18) σ p σ o (cid:19) ≤ u , (5)where the value of u , small, can be later chosen. Thisallows us to find a constraint on the maximum value, σ p,max , of the uncertainty on the participation grade asa function of the other parameters: (cid:18) σ p,max σ o (cid:19) = 2 u NM (cid:18) w p w o (cid:19) . (6) Fairness criterion (choice of u ): to ensure that the im-pact of the Performance Grade on the reliability of thetotal lab grade be negligible it suffices to choose u = 0 . σ p one order of magnitudesmaller than that of σ o . Since uncertainties are roundedto the first digit , the composition of the uncertainties ensures that the contribution coming from σ p – thus ap-pearing only on the second digit – will always be negli-gible. This way, one does not have to worry about thelarger variability of the participation grade which, given FIG. 1: Maximum value of the uncertainty on the partic-ipation grade to satisfy the fairness criterion. u = 0 . N = M , w = 0 . w p = 0 . N = M , w = 0 . w p = 0 .
05; (c) N = 2 × M , w = 0 . w p = 0 .
1; (d) N = 2 × M , w = 0 . w p = 0 .
05. The dashed,horizontal line represents the value of σ p,max for which the in-terval ± σ p,max = 1 (i.e., the uncertainty is as large as the fullgrading scale). The choice of σ o discussed at the beginning ofthis section corresponds to finding the intersections betweena vertical straight line at σ = 0 .
015 and the estimates for σ p,max . the numerous constraints (number of students, changinginstructors, etc.), may be more liable to stronger fluctu-ations.Fig. 1 shows the maximum value permitted for the un-certainty on the participation grade σ p,max , as a func-tion of the uncertainty of the ordinary grade, σ o , fordifferent values of participation grades N ( NM represent-ing the number of participation grades per lab session)and different weights w p ( w o = 1 − w p ). The horizontal,dashed line, represents the maximum possible value for σ p,max , i.e., the value for which the interval ± σ p,max covers the whole range of possible grades (i.e., from 0%to 100%). This corresponds to the limit where a Perfor-mance Grade randomly attributed (with a gaussian dis-tribution), rather than given with some pedagogical crite-rion, does not not have an impact on the accuracy of theoverall grade. Obviously, this limit is absurd since evena poorly organized and inefficient Participation Gradeaverage will do better than that, but it provides a well-defined border beyond which the curves lose meaning.We therefore take the horizontal dashed line as the cut-off for all the curves.The lowest curve (a, black online) represents the mostdifficult situation in which to fulfill the fairness criterion,since the Performance Grades is given a (relatively) largeweight (10%) and the number of Performance Gradesequals the number of Ordinary Grades ( N = M ). For anOrdinary Grading Uncertainty equalling 1% (i.e., guar-anteeing that the grade is absolutely accurate at ± σ p,max = 0 .
04 (i.e., total uncertainty ± σ o = 0 . – i.e., ± σ o ), σ p,max already grows to 0.08. Thisis a comfortable value, since it implies that globally, theuncertainty on the true value grade is not affected bythe participation grade unless its overall (3 σ ) uncertaintygrows beyond ± σ o are traced forcompleteness, but are not expected to play a serious role,since a worst-case error in evaluation of ± (cid:113) NM contribution in eq. (6). At σ o = 0 .
02 (cf. above discussion), σ p may take valuesas large as .115, thus bringing the 3 σ p,max interval to35% (thus the ± σ interval to 70%!). The other twocurves refer to smaller weights for w p = 0 .
05, which at σ o = 0 .
02 already renders the uncertainty on the partici-pation grade entirely irrelevant (its maximum value is thewhole grade interval) even when the number of participa-tion grades equals that of ordinary grades (curve (b), redonline). Eq. (6) shows that the dependence on u and on (cid:113) NM is weak (due to the square root) and therefore thecases covered by the curves of Fig. 1 are fairly complete;the changes in values coming from different choices of u , M or N will not substantially affect the expected σ p,max .Thus, we conclude that the fairness constraint can bequite easily satisfied by reasonable choices of the relativeweights for the grades, with the chosen values of u and NM . Choice of value for w p : I personally prefer choosingthe larger w p = 0 . ). Thus s/he is morelikely to focus on the straight contribution of the partici-pation grade G p to the total and a 10% contribution rep-resents more of an incentive towards good performance.Nuances exist, due to the differences coming from thegrading system (letters, percentages, numbers . . . ) andfrom local traditions, and these can be best appreciatedby the experienced Lab Supervisor, who can suitablyadapt the ideas exposed here to the local context andwisely choose the value to attribute to w p . The aim isto maximize the student’s motivation while maintaninga high degree of reliability of the final grade. V. USING THE PARTICIPATION GRADE
The grade should encourage good work and immedi-ately warn students against bad attitudes, such as wait-ing passively, trying to get the right answer by simpletrial and error – e.g., by prodding the instructor to seewhat the correct answer may be –, letting the partner(s)do the work and sitting back . . . . At the same time, itshould also be used to instill the correct working prac-tices for team collaboration. In a group where stronglevel disparities appear, one needs to push the weakerstudent not to sit back and profit from the partner(s), butalso signal the stronger student(s) to include the weakerones and help them progress. This is a very importantlesson which anyone can learn and which goes well be-yond the framework of the laboratory course. Leavingaside moral and societal considerations – which in spiteof their importance cannot necessarily be included in labinstruction – it will be clear to any student that the abil-ity of working in teams will be paramount to their futuresuccess, independently on their career choices. Thus, anegative evaluation – low participation grade – can becorrectly given to student(s) who may have personallyexcelled during lab time but have entirely disregardedand perhaps have marginalized one of more of their teammembers.In order to make full use of this grade, it is importantthat either the individual Instructor or the Lab Super-visor (depending on class sizes and organization) reservetime – preferably at the end of each session –, to dis-cuss with the students their Performance Grade, whatare the reasons which have lead to the choice that hasbeen made and which aspects of the work and attitudeneed improvement.
VI. A FEW SPECIAL CASES
One of the most difficult cases to handle in lab isthe student who is exclusively focussed on getting the right result, to guarantee for her/himself a good grade.Most instructors – typically Graduate Students or Tem-porary/Junior Faculty members – are told by the LabSupervisor in charge of the whole course not to give theanswers but let the students work it out for themselves.However, the student entirely focussed on the grade willnot easily let go. The participation grade, given at theend of each session and explained to the students, is agood way of steering grade-motivated students. Receiv-ing a bad grade (a 0% should not be discouraged since itcan be psychologically very effective, while bearing littleimpact on the final grade) right at the end of the sessionaccompanied by an explanation of the reasons for thebad grade can work wonders for instilling the right mo-tivation , particularly into those students who are grade-driven and who wouldn’t otherwise be deterred by otherarguments!Another kind of course where we have successfully usedthe Participation Grade is in advanced labs for studentsin their last year before the degree. There, studentsare required to start taking some initiative and perform(partially) independent work, on the basis of suggestionsgiven in laboratory writeups. Not all students, eventhough they may have well mastered the more technicalaspects of previous lab courses, readily turn to indepen-dent work; paradoxically, some of those best at perform-ing under guidance may have more difficulties in gainingindependence. Thus, the Participation Grade serves thepurpose of giving quick quantitative feedback all alongto more effectively guide and encourage the developmentof independence in lab.
VII. CONCLUSIONS
The Performance Grade offers the potential for guid-ance to students in their attitude towards learning lab-oratory techniques, independence, critical thinking butalso acquiring group-working practices and skills. Theevaluation of the student performance during lab sessionsis in itself a complex task, open to different and varyinginfluences which may change with Instructor, workingconditions and day-to-day events. As such, fluctuationsin the evaluation are unavoidable and may lead to worries about the volatility of this grade and its negative influ-ence on the final lab grade. I have shown that by appro-priately choosing procedures and weight coefficients, onecan limit this influence below leves which ensure that theuncertainty on the final grade remains entirely unaffectedby the larger intrinsic fluctuations of the ParticipationGrade.As a final observation, the Performance Grade will in-fluence the lab grade of each individual student, thus lift-ing the degeneracy in the group, even though this modi-fication in itself will not be very large (due to the smallweight w p ). This should be regarded as a positive contri-bution, since it is important to include in the evaluationof laboratory practices the aspects which have been listedin Tables I and II. Acknowledgments
The author acknowledges the collaboration of all stu-dents who have taken the two semesters of undergradu-ate physics laboratory course at the Universit´e de Nice-Sophia Antipolis for the past ten years and all the col-legues – young and less young , but too numerous to beindividually mentioned –, who have served as Instructors. ∗ Electronic address: [email protected] Neither qualifier covers entirely the scope of this grade.
Per-formance emphasizes the aspect related to the result (andany grade, in essence, evaluates performance): the successachieved in matching the set goals – here: the acquisitionof laboratory practices.
Participation , in the sense of activeparticipation (i.e., taking part ) stresses the active role thatthe student has to play in learning the laboratory techniquesand practices. I will use the two qualifiers interchangeablythroughout this paper. In an advanced lab it is often appropriate and recommendedto have students consult additional documentation, such asweb ressources, instrumentation manuals, textbooks, or sci-entific articles. The independence, initiative, degree of in-volvement and performance in accomplishing this tasks dur-ing lab work (if sufficient time is planned in the session) can be an integral part of the participation grade assessment. These points are summarized from unpublished guidelinesgiven to the Instructors for the courses in ExperimentalMethods in the Physics Curriculum at the Universit´e deNice-Sophia Antipolis since 2004-05. G.L. Lippi, Eur. J. Phys. , 045012 (2014). J.R. Taylor,
An introduction to uncertainty analysis: thestudy of uncertainties in physical measurements (UniversityScience Books, Mill Valley, CA, 1982). The statement holds strictly for the values of σ o which werealistically consider (cf. Fig. 1). If the significant digit in σ o were allowed to take values larger or equal to 5, then therounding procedure would increase the uncertainty on σ t .However, this case is not considered since it would requireunreasonably high uncertainties on σ oo