Research Proposal: Robust Averaging of Faces and Delusion Proneness

Background and Significance

Robust averaging is a perceptual mechanism whereby individuals decrease the influence of extreme values when integrating multiple stimuli, akin to a statistician excluding outliers to prevent skewed results (Larsen et al., 2023). Since this process is crucial for accurately interpreting emotions and complex stimuli from facial expressions toward effective social interactions, difficulties can cause significant social interactions. Robust averaging is a perceptual mechanism whereby individuals decrease the influence of extreme values when integrating multiple stimuli, akin to a statistician excluding outliers to prevent skewed results (Larsen et al., 2023). Since this process is crucial for accurately interpreting emotions and complex stimuli from facial expressions toward effective social interactions, difficulties can cause significant social interactions. Difficulties in robust averaging can lead to significant social problems due to misinterpretations of social cues. Research shows that individuals with hallucination-proneness and psychotic-like symptoms exhibit alterations in robust averaging, as they tend to weigh extreme and inlying evidence equally (Larsen et al., 2023). This has been primarily observed in low-level perception tasks, such as color perception (Larsen et al., 2023). However, it is unknown whether psychosis-proneness, including paranoia-proneness, affects robust averaging of higher-level stimuli like facial expressions. This gap in research is critical to address, as it may have significant implications for social functioning and isolation. Previous studies, such as those by Ji et al. (2018) and Ji, Pourtois, & Sweeny (2020), have explored robust averaging but did not focus specifically on schizophrenia or psychosis-risk groups. These studies suggest that perceptual integration processes can be altered in individuals with psychotic-like symptoms, indicating a broader relevance of robust averaging across different perceptual levels. Furthermore, robust averaging is crucial for interpreting group emotions and social dynamics, and difficulties in this process may lead to social isolation and impaired functioning (Haberman & Whitney, 2007). Addressing the impact of psychosis-proneness on robust averaging in social contexts could provide valuable insights into the mechanisms underlying social dysfunction in these individuals and inform interventions to improve their social outcomes.

Specific Aims and Hypotheses

Accordingly, this study aims to further explore the impact of paranoia and psychosis-proneness on robust averaging of facial expressions and its broader social implications. It also examines whether altered robust averaging affects social outcomes such as loneliness and perceived social support. The hypotheses are:

Aim 1: Test whether individuals demonstrate robust averaging for facial expressions.

• Hypothesis 1: Robust averaging for facial expressions of emotions will be observed in the general population.

Aim 2: Examine whether paranoia-proneness and psychosis-proneness are associated with alterations in robust averaging for facial expressions.

• Hypothesis 2: Individuals with high paranoia-proneness and psychosis-proneness will exhibit impaired robust averaging of facial expressions.

Aim 3: Investigate the relationship between altered robust averaging and social outcomes such as loneliness and perceived social support.

• Hypothesis 3: Altered robust averaging correlates with higher levels of loneliness and lower perceived social support.

Methods

100 Participants will include undergraduates aged 18-29 years, the peak age range for the incidence of psychotic disorders, with normal or corrected-to-normal vision, recruited from college psychology courses and through campus flyers (Kessler et al., 2007).

1) Measures

Robust Averaging Paradigm: The robust averaging paradigm will assess how participants integrate multiple stimuli, mainly focusing on facial expressions.

Paranoia-Proneness: Assessed using the Revised Green Paranoid Thoughts Scale (RGPTS; Freeman et al., 2017), an 18-item self-report questionnaire that measures ideas of persecution (e.g., "People are against me") and reference (e.g., "People are talking about me behind my back").

Delusion Proneness: Measured using the Peters Delusion Inventory (PDI; Peters et al., 2004), a 21-item self-report questionnaire designed to assess delusional ideation in the general population. Items include statements such as "Do you ever feel as if people seem to drop hints about you or say things with a double meaning?" which participants rate for their applicability.

Hallucination Proneness: Evaluated using the Cardiff Anomalous Perceptions Scale (CAPS; Bell et al., 2006), a 32-item self-report questionnaire assessing various anomalous perceptions such as changes in sensory intensity, distortion, and hallucinations (e.g., "Do you ever hear voices commenting on what you are thinking or doing?").

Loneliness: Assessed using the three-item Loneliness Scale (Hughes et al., 2004). This brief scale asks participants to rate how often they feel isolated, left out, and lacking companionship (e.g., "How often do you feel isolated from others?").

Perceived Social Support: Measured by Multidimensional Scale of Perceived Social Support (MSPSS; Zimet et al., 1988). This 12-item self-report questionnaire assesses support from family, friends, and significant others (e.g., "There is a special person who is around when I am in need").

2) Procedure

After obtaining informed consent, participants will complete a robust averaging task using PsychoPy. They will evaluate arrays of 8 facial expressions with varying emotional intensities, where mean intensity (high or low) and variance (high or low) are manipulated. Each trial includes a fixation cross (500 ms), a mask (100 ms), the stimulus array (500 ms), and the participant's response (up to 1200 ms). Participants will judge the overall emotional intensity of each array. Following the task, participants will complete the RGPTS, PDI, CAPS, three-item Loneliness Scale, and MSPSS on Qualtrics.

3) Data Analysis

Data analysis will be performed in R. Following Larsen et al. (2023), we will estimate the weight each participant gives each element in the 8-face array when making decisions by conducting logistic regressions with the element rank as the predictor (rank 1-8) and decision (“positive” or “negative”) as the outcome, on a trial-by-trial basis. Element weights (i.e., standardized beta values) will be averaged across trials for each participant.

To assess Aim 1, we will conduct two sets of analyses with these weights:

1. Quadratic Effect Analysis: We will test the quadratic effect of element rank on beta weights with a regression analysis. If participants are robust averaging, we should observe a u-shaped curve in the association between rank and weight, such that outlying ranks are given less weight.

2. Inlying vs. Outlying Weights Analysis: We will average the weight of inlying (elements 3-6) and outlying (elements 1, 2, 7, and 8) ranks. If individuals are robust averaging, inlying ranks should receive comparatively higher weights, which will be tested with a paired-sample t-test.

To assess Aims 2 and 3, we will use Pearson correlations to calculate the difference between inlying and outlying ranks as our measure of robust averaging extent and separately correlate these scores with paranoia-proneness, loneliness, and social support.

4) Sample Size Determination.

A prior study using a similar paradigm found large-magnitude robust averaging effects (ds>.86; Larsen et al., 2023). With N=100, we will have >99% power to detect the expected large-magnitude effects assessed in Aim 1 and medium-sized effects of r=.27 with 80% power for Aims 2-3 (alpha=.05, two-tailed).

Roles

I will administer the robust averaging task and questionnaires, collect and organize data from PsychoPy and Qualtrics, and perform data analysis. My faculty advisor will guide study design, data analysis techniques, and results interpretation. I will also collaborate with other research assistants for participant recruitment and data entry as needed.

Timeline

• August 2024: Submit IRB protocol

Prepare and submit all necessary documentation for IRB approval, including study protocol, consent forms, and recruitment materials. Finalize materials and address any concerns with the advisor.

• September 2024- February 2025: Test Participants

Develop recruitment materials and recruit participants through university psychology courses and campus flyers. Conduct participant testing, administer the robust averaging task and questionnaires, and ensure data quality. Monitor and troubleshoot data collection, with periodic advisor meetings for progress review.

• January 2025-March 2025: Finalize data analysis code in R.

Conduct preliminary data analysis and refine the data analysis code in R. Consult with the faculty advisor to confirm analysis techniques and ensure the code's robustness.

• March 2025-April 2025: Conduct and interpret analysis. Complete and submit the honors thesis.

Conduct final data collection and comprehensive statistical analyses. Interpret results, write, and finalize the honors thesis, incorporating feedback from the faculty advisor and peers. Submit the thesis and prepare for an oral presentation or defense if required.

References

Larsen, E. M., Jin, J., Zhang, X., Donaldson, K. R., Liew, M., Horga, G., Luhmann, C., & Mohanty, A. (2023b). Hallucination-proneness is associated with a decrease in robust averaging of perceptual evidence. Schizophrenia Bulletin, 50(1), 59–68. https://doi.org/10.1093/schbul/sbad129

Ji, L., Chen, W., Loeys, T., & Pourtois, G. (2018). Ensemble representation for multiple facial expressions: Evidence for a capacity limited perceptual process. Journal of Vision, 18(3), 17. https://doi.org/10.1167/18.3.17

Ji, L., Pourtois, G., & Sweeny, T. D. (2020). Averaging multiple facial expressions through subsampling. Visual Cognition, 28(1), 41–58. https://doi.org/10.1080/13506285.2020.1717706

Haberman, J., & Whitney, D. (2007). Rapid extraction of mean emotion and gender from sets of faces. Current Biology, 17(17), R751-R752. https://doi.org/10.1016/j.cub.2007.06.039

Kessler, R. C., Amminger, G. P., Aguilar-Gaxiola, S., Alonso, J., Lee, S., & Ustun, T. B. (2007). Age of onset of mental disorders: A review of recent literature. Current Opinion in Psychiatry, 20(4), 359-364.

Bell, V., Halligan, P. W., & Ellis, H. D. (2006). The Cardiff Anomalous Perceptions

Scale (CAPS): A new validated measure of anomalous perceptual experience. Schizophrenia Bulletin, 32(2), 366-377. https://doi.org/10.1093/schbul/sbj014

Freeman, D., Garety, P. A., Bebbington, P. E., Smith, B., Rollinson, R., Fowler, D., ... & Dunn,

G. (2017). The Revised Green et al., Paranoid Thoughts Scale (R-GPTS). Psychological

Medicine, 37(8), 1143-1151. https://doi.org/10.1017/S0033291707001638

Hughes, M. E., Waite, L. J., Hawkley, L. C., & Cacioppo, J. T. (2004). A short scale for

measuring loneliness in large surveys: Results from two population-based studies. Research on Aging, 26(6), 655-672. https://doi.org/10.1177/0164027504268574

Kessler, R. C., Amminger, G. P., Aguilar-Gaxiola, S., Alonso, J., Lee, S., & Ustun, T. B. (2007).

Age of onset of mental disorders: A review of recent literature. Current Opinion in Psychiatry, 20(4), 359-364. https://doi.org/10.1097/YCO.0b013e32816ebc8c

Peters, E. R., Joseph, S. A., & Garety, P. A. (2004). Measurement of delusional ideation in the

normal population: Introducing the PDI (Peters et al. Delusions Inventory). Schizophrenia Bulletin, 30(4), 1005-1022. https://doi.org/10.1093/oxfordjournals.schbul.a007116

Zimet, G. D., Dahlem, N. W., Zimet, S. G., & Farley, G. K. (1988). The Multidimensional Scale

of Perceived Social Support. Journal of Personality Assessment, 52(1), 30-41. https://doi.org/10.1207/s15327752jpa5201_2