The Effect of Authority Messaging on Students’ Attitudes Toward Cheating
Advisor: Mr. Aaron Betcher
Internship Location: UCSD, Psychology Department
Onsite Mentor: Dr. Tal Waltzer (PhD, UCSD)
For decades, educators and researchers have sought to understand why students cheat. Prior studies relied on self-reporting or overlooked students’ reasoning, capture students’ underlying motivations, which are essential to understanding and preventing cheating. Our paper aims to fill these gaps. Across two studies, we examined (1) what students’ social experiences surrounding academic integrity look like, and (2) how social messages might influence students’ decisions to cheat. In Study 1, we gathered naturalistic data from high schools across California using classroom observations (N = 1208 reports) and brief anonymous interviews (N = 660 interviews). Preliminary analyses of over 825 observed statements or conversations from students suggest that peer interactions around cheating are largely positive (69%) or ambiguous (24%). Teacher statements and interactions, on the other hand, were largely negative (74% of 383 statements or conversations). This shows student interactions with their peers are more pro-cheating, while their interactions with teachers are more anti-cheating. In Study 2, which is ongoing, we will test how manipulating common messages gathered from Study 1 would influence student responses in a graded quiz setting. Teachers will deliver different messages to classes (punitive: “Please don’t cheat, because you’ll get a zero and get sent to the principal’s office,” nurturing: “Please don’t cheat, because then I won’t know what material you don’t understand, and I can’t help you learn,” and a control condition: “Please don’t cheat”) and give students an opportunity to cheat on a quiz. We hypothesize that punitive reasoning will discourage cheating more compared to nurturing reasoning and the control condition. Our research makes important advancements in the field of moral development by shedding light on students’ social experiences surrounding cheating and understanding their effect on students’ tendencies to cheat.
Project Blog
Week 0: Introduction and Background
Hello everyone! I am Shirley Fan. Welcome to my first senior project blog post! Over the next few months, I will be executing experiments and performing analysis on data regarding the influence of teacher messaging on high school student perceptions and actions on academic integrity.
Cheating in general has been a long-existing problem in all levels of education and with technological advancements like the internet and ChatGPT, cheating has only become easier. Cheating is a global problem where students cut corners to get good grades but do not actually learn the material. In the long term, cheating can result in professionals who did not truly earn their credentials and do not have the skills needed for their jobs. In order to solve a problem as rampant as cheating, one first has to understand the effect different influences have on an individual’s tendency to cheat.
Being a high school student myself, I often hear talk of cheating in my peers’ conversations and constantly hear my teachers talking about academic integrity. I want to contribute to the conversation of academic integrity as I have always been interested in furthering educational access, but the essential first step to expanding educational access is encouraging students to want to learn. Prior research on academic integrity has been largely on students in higher education and relied on quantitative and self-reported data. Through this project, I aim to fill these gaps by perform an experiment in three different high schools across California to investigate the influence of teacher messaging on student perceptions and tendencies to cheat.
I am aiming to answer three main research questions in this project:
1) What are students’ perceptions of cheating?
2) How do social messages from authority figures guide students’ evaluations of cheating?
3) How do students’ evaluations of cheating guide their decisions to cheat?
This week I am working on recording some of the data we received from the first high school where we ran our experiment and throughout the next few weeks I will continue prepping material for the other high schools and analyzing the data we get from those experiments.
More details will be posted in my blog post next week. Stay tuned!
Week 1: Piloting, Data Cleansing, and Data Analysis
Hello everyone! Welcome to Week 1 of my project!
This week, I mainly worked on analyzing a data set we obtained from piloting peer messages regarding high school student views of cheating I compiled along with the lab from some data we collected last year through naturalistic observations in high school classrooms. We had respondents answer whether or not they agreed with the statements, how much more willing they would be to cheat if they heard that (-10 much less willing to 10 much more willing), and how common that statement was in their high school (0-100%).
We had a total of 58 responses. I created a coding scheme and coded each response of the 16 messages (8 cheating is okay (positive), 4 cheating is not okay (negative), and 4 ambiguous) we used, to one of four classifications: agree, disagree, mixed, and ambiguous. For willingness and commonality, I found the mean and median and compared them across the different messages. All messages had negative means and medians for willingness, but negative messages had much more negative willingness values than positive and ambiguous messages. Commonality had no linkage to the type of message used, most messages' commonality was between 30 and 50%. For the most part, negative messages had much more agreement than positive messages.
I met with the rest of my lab to discuss these results and other plans for our experimentation. Using the data analyzed above, I will use factors such as commonality or varying agreement to choose 3 negative messages and 3 positive messages for the prompted conversations portion of our study that we will execute later in the year.
As for issues, there was a slight bug in our Qualtrics survey’s connection to SONA, so many respondents were unable to get credit the first time around and did the survey multiple times. To prevent double counting their results, I performed a data cleansing by only taking the earliest responses from unique IDs.
That is all for this week. Catch you all in my next post in Week 2!
Week 2: Data, Data, and a little more Data
Hello, fellow brains!
Last week I was mostly coding the pilot data for our prompted conversations and using the criteria of most common messages and the messages with the largest effect on the willingness spectrum (most positive for positive messages, most negative for negative messages). We ended up choosing two messages per condition for the prompted conversations. Using the analyzed data, I created a handout with all the components for the interns to use.
I organized meetings with our high school interns to brief them on the prompted conversation handouts and how to report the data using the Qualtrics we made.
The rest of the week was used mostly on data entry for the experimental data from the cheating experiment we held. As of two weeks ago, we finished running the experiment in all three high schools. My task was checking all the scores of the quizzes and the amount of impossible questions they got right. There was much less cheating than we expected within the data and as of right now, it seems there was not much of a difference between the class treated with the nurturing message and the control message. I also inputted most of the survey responses of the interns and it seems the most common responses are “I do not care because it is not my business” and “I would feel bad because it isn’t fair.” It was also very common for respondents to say that cheating would be less acceptable if others were affected by it.
I also joined the broader Social Cognitive Development Lab for their review team meeting and helped out with a literature review for one of the PhD student's new study. It was interesting learning about other projects within the broader lab and contributing to them.
This week was mostly data entry, a bit more data entry, and a lot of data entry. Hope to catch you on my blog next week! Sayonara!
Week 3: Preparations, Quality Checks, & Presentations
Hello brainiacs!
This week mainly focused on preparations for prompted conversations. We met with all the interns this week to brief them on the prompted conversations, familiarize them with data submission, and created detailed timelines for all their student and teacher prompted conversation timelines. The interns have started to do and submit their prompted conversations this week and we have periodically checked in with them.
I also joined the SCDL Lab-wide meeting and watched a presentation on comparisons between self perception and perception of others’ wants in terms of forming friendships. The presentation yielded interesting results with younger children (5-7) compared to their older counterparts (8-12) the younger children percepted much less difference between what they want and while older children percepted a much larger difference. The lab-wide meetings gave me much more insight into how niche of topics psychology and development research projects can take on. A minor lesson I also learned from the presentation when Professor Heyman was giving feedback is the importance of pre-registration (uploading research questions and methods online permanently) to prevent p-hacking (going back after data collection and changing the hypothesis). I also joined the lab literature review on AI and found and reviewed literature relating to AI usage in learning and education. This opportunity allowed me to gain more experience in finding literature and one particular paper I read about the future of AI in education was particularly enlightening as it highlighted the potential for human and artificial cognition to develop together in parallel.
We also quality checked the data we entered from last week on the experiment we conducted throughout the month. We went through and checked for blanks and the accuracy of all scores and impossible questions answers inputted .
I also met up with my mentor to discuss writing a paper about the experimental study.
Catch you all next week! Ciao!
Week 4: Planning it all out
This week focused mainly on how we are going to analyze the data from our experiment. We created an outline of the different analyses we are going to perform on the experimental study data: behavioral (binary cheat/not cheat and average points cheated), attitude (using survey responses), and validation methods. Through this brainstorming process with my external advisor, I learned much of the process within the psychology field. My advisor explained to me throughout the process which analyses should be performed and which should not as they would be picked apart by reviewers and would be impossible to execute.
We also created a brief survey and wrote a pre-registration for that survey to fill in a crack we discovered in our study after meeting and discussing with the professor who leads our lab. Through that meeting I learned of the process of creating a paper and how airtight everything within the paper has to be.
I also met with an intern to fill in some blanks from the observation notes from our experimental study and checked in with our entire intern team on their progress with the prompted conversations segment of the study.
We also started our coding for our Year 2 Observational data and are planning to analyze the data for valence (positive, negative, mixed, ambiguous attitudes toward cheating), peer and teacher talk and action, and presence of a cheating event.
Overall this week involved a lot of brainstorming and planning different moving parts out.
See you all next week! Valete!!
Week 5: Analytical Analysis
Hello fellow psychological mysteries!
This week I spent most of my time breaking up our data from the observational data into single events and marking them with relevant flags for coding. Coding in psychology is going through the data and marking different data points as relevant or not to specific coding schemes—points of analysis. We are going to start agreement soon, which is when two people use the coding scheme and code a percentage of the total data set together to see if the scheme created is comprehensive or not,
My external advisor and I will also be presenting on a panel at a Social Domain Theory Conference so I have been analyzing our data and making our presentation to prepare for that talk over the week. The biggest lesson I learned from the experience is the large difference between the rigor of a talk and paper—the former being much for informal and a simple sharing of what we have been working on and interesting observations we found thus far.
We are continuing to work on prompted conversations for teachers and students and will be concluding those around mid-April. Also, I have started working on the analysis plan for our experimental study and through that have gained much more insight into the complexity of the responses we received and the analysis we will have to perform.
Wish me luck! See you all in two weeks! 再见！
Week 6: Filling in Gaps
Hello all! Welcome back from Spring Break!
This week I mainly focused on filling out a couple of gaps in my project. I met with my EA and the professor in charge of the lab to walk through the experiment and how we are going to report it. We realized there was a missing piece in our comparison, so I made a survey describing our cheating experiment scenario and published it on SONA so I could make comparisons between students' judgments of hypothetical scenarios and real-world events. I also made a version of this survey to circulate with high schoolers at the schools we worked with to get data with a population more comparable to our experimental population.
I joined our Social Cognitive Development Lab meeting and heard a presentation about AI usage and student perceptions of AI that a lab member was preparing for a talk. Through that experience I learned a lot more about how to communicate findings and what information belongs in a talk vs. belongs in a paper (talks being much more concise).
I also learned how to make agreement sheets and started setting up agreement sheets for our data analysis (to test if a coding scheme can be replicated). I also started coding for one of our experimental condition’s (nurture) appearance in our data to show how rare it is, so we can show our motivation for why we chose to investigate that particular message.
Catch you all next week! Bæ!
Week 7: Coding (but not what you think)
Hello, fellow brains!
This week I mainly worked on coding the data. Coding in psychology is not like the computer programming you are used to, it is used to classify data into different buckets. Often in coding, there is a coding scheme that outlines which scenarios are classified in which ways, and these schemes are tested using something called “agreement,” where two or more “coders” use the scheme to code a subset of the data and then compare their classifications through something called a Cohen’s kappa to see if the coding scheme is rigorous enough or not.
Right now I am working on binary coding where I am marking my data using a binary scheme (present or not present) for whether or not nurturing exists in our observational data. This will serve as a validation measure for us to explain why we chose to use the nurturing message (it is very uncommon so it is interesting to investigate). This scheme will not require a scheme as I am generally just marking anything a teacher says with remote relation to nurturing as present.
Throughout this data analysis process, I learned a lot about how to construct coding schemes as well as formulas for Excel to construct the agreement sheets. I also learned a lot about the process of data analysis within the psychological field as well as statistical analysis techniques applicable to any field.
See you all next week! Au revoir!
Week 8: Prepping it up
!sniarb wollef olleH
On Monday, at our lab meeting, I saw the many other projects that are going on within our lab and understood the breadth of projects that even a lab focused on social cognitive development can take on. We also had a guest speaker come in to talk about his PhD dissertation project that he just defended and that helped me understand more of what a PhD entails.
I also worked a lot on the presentation on our Year 1 observational data that I am going to present at a conference next week Friday, practiced it, and met with my external advisor for feedback on it. I started looking into potential journals for the paper my EA is writing and created a spreadsheet on it. I also started drafting the methods part of the paper and learned the various components that go into that section and how to organize it.
As for data, I worked on inputting more prompted conversations, and I am going to create an agreement sheet and code for agreement for our Year 2 Observational data in the next few days.
Catch you all next week! Tschüss!
Week 9: Surprising Surprises
Bzzz bzzz! What do you all think about changing answers while grading? If there was a quiz full of material you never learned, would you cheat?
Well that’s what we asked everyone using anonymous surveys over the past few weeks that described the exact scenario played out in our experiment —- the participant would have to take a graded quiz where they could not figure out the answers in any way and were given the opportunity to cheat during self-grading. Our results showed that many expressed they would cheat. But, what is even more surprising is that these same respondents expressed that teachers simply saying “Please do not cheat, because if you do, I won’t know what material you don’t understand, and then I can’t help you learn as well” instead of just “Please do not cheat” brought about statistically SIGNIFICANT differences (1-tail chi-square at .05 level of significance) between student’s answers of whether or not they would cheat (p = .035).
What is even more surprising about this finding is the fact that in our real-time high school experiments, differences between the two messages were not statistically significant (p = .51).
These extremely different p-values between the hypothetical scenario surveys and what people actually did in the situation possibly show a flaw in the popular psychology research method of using hypothetical scenario surveys. They do not always depict reality.
That was my TED talk. Catch you al next week! ลาก่อน
Week 10: And That’s a Wrap
!olleH This week I mainly focused on wrapping things up with the project. I assisted in doing data analysis of our hypothetical and experimental data and drafted and edited an abstract for it so my EA could go over it with our professor and determine how we wanted to structure the paper for this study. Through this analysis process, I realized that the usage of hypothetical scenarios — one of the most commonly used methodologies in psychology research — actually bore significantly different results from actual real-time experimentation (p = .045, p = .102, respectively at 95% confidence interval).
I also worked with my fellow lab members and crafted presentations for the different high schools that we worked with to present some of the interesting findings we had as well as some insights we had into cheating at their schools.
Last Friday, I also had the honor of presenting our observational data findings at CSUSM’s Research Conference and learned a lot about how to give a research presentation as well as how broad the field of psychology was — from work structuring to economic decisions.
Throughout pursuing this project, I learned a lot about what working in a psychology lab and research position entailed, and I was able to explore psychology as a subject in much greater detail and understand just how many aspects there are to psychology and the breadth of issues psychology covered.
That’s all for now. See you all never! !eyb dooG