Oral Sessions

Human-computer co-creativity is a part of computational creativity where a human and an AI both actively collaborate as colleagues to produce something creative such as generating music or art. This young field has huge potential to be utilized in many areas such as education, industry, and entertainment. Nowadays, AI and robots are a part of our lives. Human-AI co-creativity has the potential to transform how people interact with an AI and user perception about an AI partner. Designing and evaluating co-creative systems has many challenges due to the open-ended and improvisational nature of the interaction between the human user and the system. Interaction design is rarely discussed in the existing literature despite being a fundamental property of co-creative systems. The co-creativity literature mainly emphasizes the abilities of AI. An adequate interaction model dramatically improves the quality of the collaboration and user engagement. To investigate the trends in the interaction design of existing co-creative systems, we analyzed 73 existing co-creative systems. In most of the co-creative systems in the dataset, users use buttons, sliders, or other UI components to communicate directly with the AI. However, in most systems (n=60), the AI in co-creative systems cannot communicate back to humans, limiting their potential to be perceived as partners rather than just a tool. In a human collaboration, two-way communication leads to increased engagement. This research presents a user study with 38 participants to explore the impact of AI-to-human communication on user perception, collaborative experience and user engagement in co-creative systems. To guide the study design and data analysis, we formulated three research questions: How does AI-to-human communication influence collaborative experience in human-AI co-creation? How does AI-to-human communication influence user engagement in human-AI co-creation? How does AI-to-human communication influence user perception of the co-creative AI agent? The study involves user interaction with two prototypes of a co-creative system that contributes design inspirations during a design task, Creative PenPal. One prototype utilizes only human-to-AI interaction (baseline) and another uses two-way communication between humans and AI, including AI-to-human communication via speech, text and a virtual AI character. All the participants tested both of the prototypes with two design tasks while we counterbalanced the order of the tasks. After each design task, the participants completed a survey to evaluate the system. Finally, the study ended with a follow-up semi-structured interview to collect qualitative data about the user engagement and overall experience with the AI. The interview and survey results show improved collaborative experience and user engagement with the system incorporating AI-to human communication (two-way communication) as the co-creative AI is perceived as a collaborative partner. AI-to-human communication also positively changes user perception of co-creative AI as users perceive it as more intelligent and reliable. This research leads to new insights about designing effective human-AI co-creative systems and lays the groundwork for future studies. Additionally, insights from this research can be transferred to other fields that involve human-AI interaction and collaboration, such as education, entertainment, and professional work.

As data drive our world and decisions, there has been a consistent call for statistical literacy (SL) and data science in K-12 mathematics by researchers and professional organizations over the last few decades (e.g., Craig, 2018; Gal, 2002; Gewertz, 2020 National Governors Association Center for Best Practice & Council of Chief State School Officers, 2010; Wallman, 1993), with a recent shift toward stressing statistics and data science in K-12, including from a consumer orientation (e.g., Gewertz, 2020; North Carolina Department of Public Instruction, 2020). These calls have been amplified by the widespread misinformation of the pandemic and the need to consume statistical information effectively (e.g., Engledowl & Weiland, 2021; Watson & Callingham, 2020). From the consumer orientation, SL is conceptualized as the skills needed to effectively make sense of statistical messages in the real world. Despite this growing emphasis on SL and data science, there is a disconnect between the skills needed to effectively consume real-world statistical messages and what is taught in schools (e.g., Nicholson et al., 2019), illuminating the need for articulating Critical Statistical Literacy Habits of Mind (CSLHM) for making sense of such messages, particularly using a critical lens. The purpose of this study is to explore the similarities and differences of how people enact CSLHM when presented with data representations from the media. This study draws on Gal’s (2002) SL model and Weiland’s (2017) framing of critical statistical literacy (CSL). Gal’s work provided a solid starting point for describing the habits of mind needed to enact CSL. Weiland’s CSL framing explicitly stressed the importance of placing emphasis on sociopolitical inequity and the actions needed to disrupt and dismantle such inequity. Specifically, I use the CSLHM (Authors, 2019, under review) which operationalized the habits of mind needed to enact CSL. The CSLHM framework has seven components: questioning sample size and methods, recognizing appropriate statistics, desiring additional information, acknowledging alternate explanations, recognition of one’s own sociopolitical consciousness, employing active citizenry, and acknowledging ethical considerations. In this instrumental multiple case study (Yin, 2018) there were four cases; each defined as a group of individuals with similar backgrounds with respect to statistical self-efficacy (Finney & Schraw, 2003) and critical consciousness (Diemer et al., 2020). There were five participants in each case. Each participant took part in a semi-structured task-based interview (Goldin, 2000) in which they were asked to make sense of data representations shared on Twitter. All tweets included a static or interactive graph and were related to social justice issues. Data were first coded using the CSLHM framework (DeCuir-Gunby et al., 2011). Next, I looked for emerging themes across the codes to write up case summaries. Preliminary findings suggest that statistical knowledge, statistical self-efficacy, and critical consciousness appear to influence the enactment of CLSHM. Given that there is a difference in CSLHM enactment, future research should endeavor to determine how to best support students and teachers in developing CSLHM.

Bacterial meningitis is associated with devastating inflammation within the central nervous system (CNS), mediated by both resident brain cells and the recruitment of inflammatory leukocytes, including neutrophils. However, the molecular mechanisms underlying the initiation and progression of this inflammatory disorder are poorly understood. In our previous studies, we have demonstrated that the neuropeptide substance P is capable of exacerbating bacteria-associated inflammation within the CNS. In the present study, we have begun to determine whether substance P can augment CNS cell-mediated recruitment and/or activation of neutrophils in response to bacterial challenge. In the present study, we have employed human microglia (hμglia) and neutrophil (HL60) cell lines to investigate the ability of resident CNS cells to recruit and activate neutrophils following challenge with Neisseria meningitidis, a major causative agent of bacterial meningitis. Furthermore, we have begun to investigate the effects of the neuropeptide substance P on human glial and neutrophil responses to this bacterial challenge. Utilizing fluorescence-activated cell sorting (FACS), we confirmed the expression of mature human neutrophil markers on the surface of differentiated HL60 cells. In addition, we have used specific capture enzyme-linked immunosorbent assays (ELISAs) to assess the inflammatory responses of both cell types to N. meningitidis and/or bacterial products in the presence or absence of substance P. Finally, we have initiated migration assays employing Transwell tissue culture plates to assess the ability of microglia-derived chemotactic factors to recruit neutrophils. We demonstrate that human microglia-like cells produce inflammatory mediators in response to the bacterial products lipopolysaccharide and Pam3Cys-Ser-(Lys)4 trihydrochloride, and whole viable N. meningitidis infection. Such responses include the production of the potent inflammatory cytokine interleukin-6 (IL-6) and the key neutrophil chemoattractant IL-8. Importantly, we have found that differentiated human neutrophils (as determined by relative expression of the cell surface markers CD11b, CD35, and CD71) constitutively express the specific receptor for substance P (neurokinin-1 receptor; NK-1R). Furthermore, we have determined that substance P can augment the expression of IL-8 by neutrophil-like cells stimulated with bacterial products. These preliminary studies indicate that human glial cells respond to bacterial components and N. meningitidis to produce key inflammatory mediators, including the neutrophil attracting chemokine, IL-8. Moreover, differentiated HL60 human neutrophil-like cells constitutively express NK-1R, the receptor for substance P, and this neuropeptide can exacerbate their responses to bacterial components. Collectively, these studies support the notion that substance P can exacerbate bacteria-induced glial responses that serve to recruit neutrophils to the CNS and/or augment their activation upon their arrival. Ongoing studies employing Transwell migration assays are being conducted to determine whether substance P can directly or indirectly augment the ability of N. meningitidis-challenged microglia to induce neutrophil migration.

The third leading cause of cancer-related deaths in the United States is pancreatic cancer, >95% of which is Pancreatic Ductal Adenocarcinoma or PDAC. In the last 40 years there has been no improvement in therapy for PDAC. STATEMENT OF PURPOSE: MUC1 is a transmembrane protein with a large number of sugar residues, and is expressed on normal glandular epithelial cells. In PDAC, MUC1 is overexpressed and has less and aberrant sugar residues attached to it and is designated tumor-associated MUC1 (tMUC1). Over 80% of human PDACs express tMUC1. In an NCI initiated study, out of 75 tumor antigens, MUC1 was ranked the second most targetable antigen to develop cancer vaccines. Furthermore, the 72 amino acid cytoplasmic tail of MUC1 (MUC1 CT) is reported to aid in oncogenic signaling leading to tumor progression and metastasis by blocking cell death. A novel monoclonal antibody, TAB004, has been developed specifically against tMUC1. TAB004 detects tMUC1 with a high rate of specificity and sensitivity and spares recognition of normal epithelial MUC1. Treatment with TAB004 curbs PDAC cell survival by blocking MUC1 CT associated oncogenic signaling and renders the cells more susceptible to standard chemotherapy drugs. Several human PDAC cell lines based on their tMUC1 expression were grown in media with heat-inactivated serum to ensure that it is devoid of complement proteins. Cells were treated with various concentrations of TAB004 antibody, control IgG, 5-FU, Paclitaxel (PTX), or Gemcitabine and the IC50 was determined using colony forming assay. Once IC50 was determined, cells were treated with combinations of TAB004 and the drugs. CFPAC and MiaPaca2 cells were treated with IgG or TAB004 for 24 hours and transcriptomics analysis was performed. To determine apoptosis, treated cells were stained with Annexin V-FITC and PI, and analyzed by flow cytometry and expression of cleaved Caspases by western blot. Data analysis was performed using GraphPad Prism 9.1 and a p value of <0.05 was considered significant. CFPAC cells were injected in nude mice and treated weekly with IgG or TAB004, and tumor growth was monitored. TAB004 treatment alone inhibited the colony forming ability of most of the human PDA cell lines. In addition, when combined with Gemcitabine, PTX, or 5-FU, TAB004 significantly increased anti-tumor efficacy of the drugs. Transcriptomics data revealed a number of differentially expressed genes in IgG vs TAB004-treated cells which are possibly responsible for inhibition of colony forming potential in these cells. Annexin V-FITC and PI staining and increased cleaved Caspases confirmed induction of apoptosis in these cells. TAB004 treatment significantly slowed tumor growth and reduced tumor burden in nude mice compared to IgG treatment. TAB004 inhibits colony forming potential of PDAC cells and enhances the efficacy of chemotherapeutic drugs. TAB004 phosphorylates MUC1 CT and activates apoptosis in PDAC cells. TAB004 slows tumor growth in vivo. Further analysis of the transcriptomics data will determine the mechanism of TAB004 induced apoptosis and sensitization of PDAC cells to drug-induced killing.

Topology Optimization (TO) is a powerful computational design method for automatically generating a structural layout to determine the optimal material layout in a design domain with maximum performance under relevant design specifications. In recent years, considerable research efforts have been made in the advancement of TO procedures. Although structural TO has great potential for creating innovative structural designs without prior knowledge, it is a time-consuming task. In this study, we investigate the application of neural networks and convolution neural networks to conduct TO directly from the finite element solver. In this approach, a neural network is used to represent the density field function independent of finite element mesh. A Fourier space projection has been implemented within the machine learning model to control the minimum and maximum length scales to meet the manufacturing and other functional requirements. We have adapted the high-performance Google automatic differentiation library JAX to build an end-to-end differentiable network model. The sensitivity computations are automated by using the built-in backpropagation functionality in JAX. The performance of the proposed framework is demonstrated by solving several elastic and thermoelastic compliance minimization problems and comparing the results with the optimized structures obtained from other optimization techniques.

Cathemerality is a unique and flexible activity pattern found mostly in lemurs. Cathemeral lemurs exhibit relatively evenly distributed activity across the 24-hour period. Although some species are broadly recognized as being cathemeral, other species have activity patterns that are still undetermined. The activity patterns of animals are largely influenced by environmental cues such as temperature, humidity, rainfall, and lunar illumination. However, the ways that animals respond to these cues are dependent on their own morphology and adaptations to a particular niche. This study examined the activity patterns of three species (Eulemur mongoz, Lemur catta, and Varecia rubra) of semi-free ranging lemurs living in the same 5-acre forest enclosure at the Lemur Conservation Foundation in Myakka City, Florida. Cross-species studies on captive and semi-free ranging populations can control for environmental variables and allow for inferences to be made about the driving forces of these activity patterns. Two individuals from each of the three species were fitted with accelerometers and activity was recorded in one-minute intervals for 45 days. Two individuals from each of the three species were fitted with accelerometers and activity was recorded in one-minute intervals for 45 days. Hourly temperature, hourly humidity, daily rainfall, day length, and nightly illumination index (NII) were also recorded. The mean daily activity divided by the mean nightly activity was obtained for each individual (DN ratio) and used in GLMM, ANOVA, and ANCOVA analyses. Daily mean activity and nightly mean activity were also used for some analyses. Due to the failure of an accelerometer and the introduction of a newborn into the V. rubra group, this species was excluded from cross-species analyses. Results suggest that daily temperature, rainfall, and humidity have little influence on semi-free ranging lemur activity when compared to their wild counterparts. However, day length and NII had a significant influence on E. mongoz activity and some influence on L. catta activity, although it was not significant. ANCOVA analyses controlling for the influence of day length and NII revealed that there was a significant difference in activity patterns between species, but that temperature, humidity, and rainfall still had no significant influence. Overall, mongoose lemurs were found to exhibit a more cathemeral activity pattern (DN ratio = 1.13) than ring-tailed lemurs (DN ratio = 1.67) and were found to exhibit more nocturnal activity than diurnal activity when NII was high. An understanding of the behavioral flexibility of lemurs and the evolutionary context of cathemeral behavior would allow us to make inferences about the impending impacts of climate change and other anthropogenic disturbances and provide us with insight on how we could mitigate or minimize these impacts on endangered lemur species. This knowledge also helps us to make informed decisions about relocations, breeding programs, animal welfare, and the release of a previously captive animal. Primates provide many ecosystem services to humans and are important members of their ecological communities. However, they face many threats and populations worldwide are continuing to decline despite current conservation efforts.