The Relationship Between Performance and Trust in AI in E-Finance.

Artificial intelligence (AI) is fundamentally changing how people work in nearly every field, including online finance. However, our ability to interact with AI is moderated by factors such as performance, complexity, and trust. The work presented in this study analyzes the effect of performance on trust in a robo-advisor (AI which assists in managing investments) through an empirical investment simulation. Results show that for applications where humans and AI have comparable capabilities, the difference in performance (between the human and AI) is a moderate indicator of change in trust; however, human or AI performance individually were weak indicators. Additionally, results indicate that biases typically seen in human-human interactions may also occur in human-AI interactions when AI transparency is low.

Data on the Human Versus artificial intelligence process management experiment.

Human subject experiments are performed to evaluate the influence of artificial intelligence (AI) process management on human design teams solving a complex engineering problem and compare that to the influence of human process management. Participants are grouped into teams of five individuals and asked to generate a drone fleet and plan routes to deliver parcels to a given customer market. The teams are placed under the guidance of either a human or an AI external process manager. Halfway through the experiment, the customer market is changed unexpectedly, requiring teams to adjust their strategy. During the experiment, participants can create, evaluate, share their drone designs and delivery routes, and communicate with their team through a text chat tool using a collaborative research platform called HyForm. The research platform collects step-by-step logs of the actions made by and communication amongst participants in both the design team's roles and the process managers. This article presents the data sets collected for 171 participants assigned to 31 design teams, 15 teams under the guidance of an AI agent (5 participants), and 16 teams under the guidance of a human manager (6 participants). These data sets can be used for data-driven design, behavioral analyses, sequence-based analyses, and natural language processing.

Using workplace thriving theory to investigate first-year engineering students' abilities to thrive during the transition to online learning due to COVID-19.

Background: During the onset of the COVID-19 crisis, universities rapidly pivoted to online formats and were often unable to adhere to the best practices of online learning highlighted in prior literature. It is well documented that a variety of barriers impeded "normal" educational practices. Purpose/Hypothesis: The purpose of this paper is to investigate the perceptions of first-year engineering students enrolled in an introductory engineering design course during the rapid transition to online working environments. We view students' perceptions through the theoretical lens of workplace thriving theory, a framework that allowed us to capture aspects of education required for students to thrive in non-optimum learning settings. Design/Method: This research employed semi-structured interview methods with 13 students enrolled in an introductory engineering design course that relies on project-based team learning. We analyzed interview transcripts using thematic analysis through an abductive approach and made interpretations through workplace thriving theory. Results: Results indicated that students' abilities to thrive are related to four intersecting themes that demonstrate how workplace thriving theory manifests in this unanticipated online setting. These themes demonstrate elements that must be optimized for students to thrive in settings such as this: relationships with others, building and sharing knowledge through interactions, perceptions of experiential learning, and individual behaviors. Conclusion: Our research, viewed through workplace thriving theory, highlights the mechanisms by which students tried to succeed in suboptimal environments. While not all our participants showed evidence of thriving, the factors required for thriving point to opportunities to harness these same factors in in-person instruction environments.

Data on the design and operation of drones by both individuals and teams.

Human subject experiments are performed to assess the impact of artificial intelligence (AI) agents on distributed human design teams and individual human designers. In the team experiment, participants in teams of six develop and operate a drone fleet to deliver parcels routed to multiple locations of a target market. Among the design teams in the experiment, half of the design teams are human-only teams with no available AI agent. The other half of the design teams, designated as hybrid teams, have drone design and operation AI agents to advise them. Halfway through the team experiment, team structure is changed unexpectedly, requiring participants to adapt to the change. In the individual experiment, participants develop drones based on given design specifications, either on their own or with the availability of a drone design AI agent to advise them. During these experiments, participants configure, test, and share their designs and communicate with their teammates through an online research platform. The platform collects a step-by-step log of the actions made by participants. This article contains data sets collected from 44 teams (264 participants) in the team experiment and 73 participants in the individual experiment. These data sets can be used for behavioral analysis, sequence-based analysis, and natural language processing.

Image-based data on strain fields of microstructures with porosity defects.

The present article provides a compilation of microstructures and respective strain fields expressed by them during elastic loading. These microstructures were synthesized in Abaqus Standard software and their strain fields were modelled using Abaqus based static implicit analysis. The Python Development Environment (PDE) in Abaqus was used. These microstructures were subjected to uniform displacement boundary condition to obtain strain fields in the plane-strain mode. The purpose of the generating this data was to test the efficacy of convolutional neural networks (CNNs) in predicting strain fields. This raw data consisting of microstructure and their strain fields was converted to images using MATLAB as two dimensional arrays with each pixel denoting value to be used as input for training the CNN. This processed data in the form of images can be potentially used in deep learning or data science methodologies to perform finite element simulations.

Image collection of 3D-printed prototypes and non-3D-printed prototypes.

Image processing refers to the use of computer algorithms to manipulate and enhance digital images to improve their quality or to make them more suitable for tasks such as classification. Common benchmarking datasets in this field include the imagenet, CIFAR-100, and MNIST datasets. This dataset is a collection of images that are particularly relevant to engineering and design, consisting of two categories: 3D-printed prototypes, and non-3D-printed prototypes This data was collected through a hybrid approach that entailed both web scraping and direct collection from engineering labs and workspaces at Penn State University. The initial data was then augmented using several data augmentation techniques including rotation, noise, blur, and color shifting. This dataset is potentially useful to train image classification algorithms or attentional mapping approaches. This data can be used either by itself or used to bolster an existing image classification dataset.

An immunoturbidimetric assay for bovine haptoglobin.

In cattle, the serum protein haptoglobin (Hp) is a major acute phase protein (APP) that rises in concentration over a thousand fold following stimulation by pro-inflammatory cytokines. As such, this APP is a valuable biomarker for infection, inflammation and trauma in cattle. The assay for bovine Hp is becoming more commonplace in clinical pathology and in experimental studies when a biomarker of innate immunity is required. The most widely used assay for Hp utilises its binding to haemoglobin (Hp-Hb binding assay), which at low pH enables the preservation of the native peroxidase activity in the haemoglobin. This assay is used for all species, including species such as dog, cat and pig where the level of Hp is higher in healthy animals of these species than in healthy cattle, and therefore a bovine-specific immunoassay that can be automated would be desirable. Thus, a novel-automated species-specific immunoturbidimetric (IT) assay has been developed. Validation studies showed intra- and inter-assay CVs of below 5% and 9% respectively and a recovery of 99% from samples spiked with bovine Hp and a limit of quantification of 0.033 g/L. The assay is not affected by icterus or lipaemia but had moderate interference from haemoglobin and showed a significant correlation with the Hp-Hb binding assay. This novel IT assay for bovine Hp will allow automated analysis of this important bovine APP to identify changes in the Hp concentration not detectable by current Hp-Hb binding assays. It will enable the incorporation of this assay into herd health assessments, animal welfare analysis and for bovine medicine and research.

Data on the design of truss structures by teams of engineering students.

This experiment was conducted in order to compare different approaches that human teams use to solve design problems that change dynamically during solving. Specifically, study participants were given the task of designing a truss structure (similar to a bridge spanning a chasm) in teams of three. At two points during design, the problem statement was changed unexpectedly, requiring participants to adapt. Two conditions were given different initial problem representations. During the study, every participant had access to a computer interface that allowed them to construct, test, and share solutions. The interface also made it possible to collect a step-by-step log of the actions made by participants during the study. This article contains data collected from 48 participants (16 teams). This data has been used previously in behavioral analyses, sequence-based analysis, and development of computational models.

Data on the configuration design of internet-connected home cooling systems by engineering students.

This experiment was carried out to record the step-by-step actions that humans take in solving a configuration design problem, either in small teams or individually. Specifically, study participants were tasked with configuring an internet-connected system of products to maintain temperature within a home, subject to cost constraints. Every participant was given access to a computer-based design interface that allowed them to construct and assess solutions. The interface was also used to record the data that is presented here. In total, data was collected for 68 participants, and each participant was allowed to perform 50 design actions in solving the configuration design problem. Major results based on the data presented here have been reported separately, including initial behavioral analysis (McComb et al.) [1], [2] and design pattern assessments via Markovian modeling (McComb et al., 2017; McComb et al., 2017) [3], [4].