Student Research in Psychology

Research Abstracts

• Pre-to-Post Data Analysis of ShineOn! Curriculum on Sixth Graders’ Social-Emotional Skills
  • Presenter: Teddy DeFayette
  • Faculty mentor: Dr. Chatlos

  Pre to post data analysis of ShineOn! curriculum on sixth graders’ social-emotional skills

  In a 2010 study, Bierman et al. showed how social emotional learning programs could increase prosocial behavior and academic outcomes, as well as reduce aggressive behaviors in children. The Shine On! program developed at SUNY Plattsburgh is a novel SEL intervention program being implemented in school districts across Clinton, Essex, and Franklin counties. The current study compares the pre to post test results of sixth grade students at Peru Central School. Students completed the Delaware Social Emotional Competencies Scale R-2 (DSECS-R2; Bear et al., 2021) before and after participating in the Shine On! intervention. After completion, students’ pre and post-scores on the five social-emotional competencies; self-management, social awareness, self-awareness, relationship skills, and responsible decision-making, as well as their total SEL scores, were examined. No significant change was discovered between pre and post SEL scores after the implementation of the Shine On! curriculum. Future research looking at different measures such as teachers’ observations of students’ post-intervention outcomes is needed.
• Cloze Tests Reveal if AI was Used to Write Your Essay, But Not if it was Used to Fix Your Grammar
  • Presenter: Thomas Konnick
  • Faculty mentor: Dr. Mansfield

  Cloze tests reveal if AI was used to write your essay, but not if it was used to fix your grammar

  In a cloze test, every fifth word of an essay is deleted and the student is required to write in the missing words. Our previous work suggests that low accuracy on a cloze test may signal that the student relied on AI to write the essay. This is because it is difficult for students to fill in the missing words when they are unfamiliar with the AI’s style and word choices. However, our previous study used simple writing prompts that did not reflect typical college-level assignments. To increase external validity, we re-examined this effect using essays students had written earlier in the semester as part of an exam. Each student completed three cloze tests: one based on their original essay, one based on an AI-revised version of that essay (to fix grammar and spelling errors), and one created entirely by AI. Repeated-measures ANOVA, F_2,33=54.80, p<.001, showed that students scored significantly lower on cloze tests based on fully AI-generated passages (M=38.0%, SE=3.35) than on tests based on their own writing (M=70.0%, SE=3.37) or AI-revised versions of their writing (M=70.6%, SE=3.37). These findings suggest that: (a) cloze tests can help distinguish AI-generated text from student-written work; and (b) using AI to improve grammar and clarity does not reduce students’ familiarity — or ownership — of their writing.

• Same Sentences, Different Senses: How Input Modality Shapes Comprehension & Production
  • Presenter: Faline Yang
  • Faculty mentor: Dr. Frinsel

  Same sentences, different senses: How input modality shapes comprehension and production

  Language learners are faced with an incredibly complex task whereby they must develop an understanding of the continuous stream of sounds by recognizing individual words, figuring out their meaning, determining their grammatical and thematic roles, and piecing all of this together in real-time into an understanding of who does what to whom. Research has shown that learners can use cross-situational statistics to learn simple sentences. However, most studies have focused primarily on the learning of a language through listening (auditory) rather than reading (orthographic), while testing comprehension (i.e., recognition) rather than production (i.e., recall). This study investigates how input modality (auditory vs. orthographic) affects the cross-situational learning of simple sentences, assessing both comprehension and production across two experiments. During the training phase, participants viewed two scenes on the screen and selected the scene they believed matched the auditory or written sentence. In Experiment 1 (comprehension), participants were subsequently tested with single scenes paired with either auditory or written sentences and indicated whether the sentence matched the scene. Participants trained in either modality showed significant learning. Performance did not differ between training modalities (p = 0.4), and participants generalized fully across modalities: being trained in one modality (e.g., auditory) did not provide an advantage when tested in the same modality. These results suggest that comprehension in cross-situational learning is robust to input modality, with learners achieving comparable understanding regardless of how the language input was delivered. Experiment 2 (production) is currently underway to determine whether these effects extend to recall, providing further insight into the role of input modality and output in cross-situational language learning.

• Time to Control: Linking Effort, Agency, and Time Perception
  • Presenters: Meghan Cox, Teddy Defayette, & Myles Maldonado-McCabe
  • Faculty mentor: Dr. Potts

  Time to control: Linking effort, agency, and time perception

  A core feature of intentional behavior is the sense of agency—the feeling of control over one’s actions and their outcomes. Temporal binding, the compression of perceived action–outcome intervals, is often treated as an implicit measure of agency. However, recent evidence questions its alignment with explicit reports of control. Here, we examine a new implicit measure: errors in perceived action durations under varying levels of effort. Participants completed a virtual aiming task requiring rapid, accurate movements. We manipulated action difficulty by varying target sizes. In separate conditions, participants either provided explicit control ratings after each action or estimated the duration of their action by clicking and holding the mouse. Preliminary results reveal similar patterns across both measures: more difficult actions are underestimated in duration and elicit lower felt control, whereas easier actions are overestimated and associated with higher control ratings. Ongoing analyses will examine whether action-duration errors predict explicit agency at the individual-participant level.

• Tour It, Pledge It, Workout: Promoting Student Fitness Center Use
  • Presenters: Graham Richard, Julianna Digiorgio, Elysia Saba, and Nate McCarthy
  • Faculty mentor: Dr. Bator

  Tour it, pledge it, workout: Promoting student fitness center use

  The current experiment examined the impact of a public commitment and an in-person tour to increase intentions to use the campus Fitness Center. We predicted that participants who made a public commitment and completed a tour would report higher intent to use the Fitness Center than those in the waitlist condition. Participants were randomly assigned to one of three experimental conditions. The commitment condition received a tour of the Fitness Center, provided input on key areas, completed and posted a commitment form, and answered an exit ticket rating intent to use the Fitness Center in the next seven days using a scale from one (very low) to nine (very high). Participants in the tour only condition had the same experiences but did not complete or post a commitment form. Participants in the waitlist condition used a Google Form to rate intent to use the Fitness Center and scheduled a tour for later in the semester. A one factor ANOVA revealed a significant effect for the commitment manipulation, F(2, 57) = 5.00, p = .01. A post hoc LSD test was conducted. This test revealed that the commitment condition and tour only condition both had significantly higher intentions than the waitlist control condition. These findings suggest that even a simple tour can be an effective strategy to increase students’ intent to use the Fitness Center.

• Students’ Perspectives on Using ChatGPT as a Therapist
  • Presenter: Erkam Arklan
  • Faculty mentor: Dr. Mansfield

  Students’ perspectives on using ChatGPT as a therapist

  The wide availability of large language models (such as ChatGPT), along with their ability to generate conversational responses that align with therapeutic approaches, raises the possibility that they could serve as a convenient, on-demand alternative to traditional appointments with a therapist. To understand how students view this possibility, we created transcripts of two 5-min counseling sessions: one adapted from real counseling sessions (extracted from publicly available training materials), and one generated by ChatGPT. Both feature a therapist employing cognitive-behavioral strategies to help a client with anxiety. Text-to-speech software was used to convert these transcripts into audio recordings. Participants will listen to these recordings, randomly assigned to two conditions: one group will be told which session was created by ChatGPT, while the other group will receive no information about the sessions. After listening, students will rate how successful each session seemed and indicate whether they would be willing to work with that therapist. Although data collection is ongoing, we expect that students who are unaware of which session is AI-generated will evaluate the two sessions similarly, whereas students who are informed about the AI-generated session will show lower confidence in that session and reduced willingness to seek therapy from that counselor. Hopefully, this study will help us understand students’ openness to technology-based mental health support.

• It’s More Than Just Participation: Student Involvement & Well-Being
  • Presenter: Emilia Kane
  • Faculty mentor: Dr. Christy

  It’s more than just participation: student involvement and well-being

  Undergraduate students typically experience the most amount of freedom for the first time during the first few years of college. Adjustment to new experiences varies for each student due to a number of reasons; personality, finances and general wellbeing. We want to explore if students who are more involved tend to be more adjusted compared to individuals who are not involved. Different veins of involvement such as club, organization and sports membership, TA/CA positions and employment on and off campus are examined, as well as the degree of involvement. While data collection is ongoing, we predict students who are more involved will experience greater wellbeing socially, emotionally and academically compared to non-involved students.

• Double Trouble: Learning Homonyms & Synonyms Through Cross-Situational Word Learning
  • Presenter: Emily McArdle
  • Faculty mentor: Dr. Frinsel

  Double trouble: Learning homonyms and synonyms through cross-situational word learning

  A key task for language learners is to determine which referents reliably co-occur with specific words. Numerous studies have shown that adults and infants can learn word–referent mappings (1:1) from cross-situational statistics—the repeated concurrence of words and referents across multiple exposures and contexts. However, natural language learning often involves greater complexity, such as homonyms, where one word maps to multiple referents (1:2), and synonyms, where multiple words map to the same referent (2:1). In our first experiment, we investigated how learning differs across these three conditions. Participants were randomly assigned to one of the three groups and learned word–referent mappings over several training blocks followed by a test phase. Results showed that overall performance was above chance in all groups. One-to-one form–meaning mappings were learned significantly more efficiently than mappings involving any type of ambiguity. The specific direction of ambiguity mattered less than the mere presence of competition: in the absence of feedback, only the 1:1 condition showed continued improvement across blocks, whereas performance in the 2:1 and 1:2 conditions plateaued and did not differ from one another. The continued difficulty with homonym and synonym mappings points to more than just slower learning; the plateau suggests that learners may require additional contextual cues to resolve competition. Therefore, in our second study, we investigated whether explicit feedback can help resolve this ambiguity. Preliminary evidence indicated that feedback provides the critical information needed to disambiguate competing mappings in both homonym and synonym conditions. Together, these findings suggest that while adults can handle some ambiguity in word learning, consistent 1:1 mappings are acquired more readily than mappings involving multiple referents (homonyms) or shared referents (synonyms), especially in the absence of feedback. When ambiguity increases, feedback appears to play a crucial role in supporting learners by enabling them to overcome mapping competition.