KJKeith S. Jones

Keith S. Jones

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Keith S. Jones has a total of 49 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    ACT: A Conceptual Framework for Understanding and Controlling Trustworthiness in Controllable AI

    by Venkata Sai Prathyush Turaga, Saroj Gopali, Peter Kieseberg, Timea Pahi, Simon Tjoa, Keith S. Jones, Akbar Siami Namin

    With the rapid advancement of AI and foundation models, several research directions have emerged which are active research focuses contributing to advancing and stabilizing the next generation of cutting-edge AI technologies. Instances of such research efforts are being explored through various concepts such as AI Alignment, Explainable AI and Transparency, AI Compliance, AI Ethics, Constitutional AI, and notably Controllable AI. Although the needs for all these research concepts and focuses are well-justified, there is little known how these areas contribute to trustworthiness in AI and whether there is any overlap or missing research gaps that need to be explored for addressing the hard-core research problems and concerns in trustworthy AI. Furthermore, it is also primitive to understand how trustworthiness can be addressed in the context of “agentic AI”, and more notably “controllable AI” where LLM-powered agents collaboratively reason and decide about workflows and tasks without human’s intervention. The key idea is that while “agentic AI” offers sophisticated technology and paradigm for the next generations of computing, the trustworthiness of such fascinating technology is still in its infancy, and therefore controllability of such agentic AI technology needs to be explored. This conceptual article intends to explore and view the research focuses in AI (e.g., AI Alignment, Explainability) in accordance with AI trustworthiness. More specifically, the work compares and contrasts these conceptual concepts with respect to the NIST Trustworthy AI framework that embodies seven criteria that include safety, security, explainability, privacy, fairness, validity, and accountability. The article provides a conceptual mapping between each of these concepts and the trustworthiness criteria listed in the NIST trustworthy framework. The paper introduces ACT, an Active framework for controlling trustworthiness criteria in AI systems.

  • Multimodal Deception Detection Using Linguistic and Acoustic Features

    by Tien Nguyen, Faranak Abri, Akbar Siami Namin, Keith S. Jones

  • Open Access

    The Applicability of LLMs in Generating Textual Samples for Analysis of Imbalanced Datasets

    by Saroj Gopali, Faranak Abri, Akbar Siami Namin, Keith S. Jones

    In machine learning class imbalance is a pressing issue, where the model is biased towards the majority classes and underperforms in the minority classes. In textual data, the natural language processing (NLP) model bias significantly reduces overall accuracy, along with poor performance in minority classes. This paper investigates and compares the performance of transformer-based models, such as Multi-head Attention with the data levels and algorithmic levels approaches and BERT (Bidirectional Encoder Representations from Transformers) with LLM-based data augmentation. The research utilized the approaches, such as Random Over Sampler, Synthetic Minority Over-sampling Technique (SMOTE), SMOTEENN, data augmentation at word level, class weights, L2 regularization and leveraging GPT-3.5-Turbo’s for data augmentation to create additional data samples in imbalance dataset. The results from the experiment demonstrate that the LLM-based data augmentation with Multi-head Attention and BERT in the Myers-Briggs Type Indicator (MBTI) dataset (a highly skewed dataset) achieves the highest precision, recall and F1 score of 0.76 across terms. It indicates that the LLM-based data augmentation has significant improvements in dealing with class imbalance and improves the model’s accuracy in minority class types in the MBTI dataset.


  • Investigating What Variables People Pick Up When Perceiving Other People’s Maximum Vertical One Degree-of-Freedom Reach Heights to Inform the Design of Assistive Robots

    by Keith S. Jones, Nicholas A. Garcia, Dana L. Wilder

    We aim to design assistive robots that perceive people’s affordances in ways that are similar to how people perceive other people’s affordances. Toward that end, two experiments investigated what variables people pick up when perceiving actors’ maximum vertical one degree-of-freedom reach heights. In Experiment 1, point-light displays depicted actors who moved, were either tall or short, and had markers placed on either their whole body (head, shoulders, elbows, wrists, hands, hips, knees, & ankles), upper body (head, shoulders, elbows, wrists, & hands), lower body (hips, knees, & ankles), or ankles. Participants instructed the experimenter to adjust an object’s height so that it was just reachable. Experiment 2 was identical except actors moved or were still. In both experiments, judgment error for the Full Body condition was not significantly different from that for the Upper or Lower Body conditions, but was significantly different from that for the Ankles condition. In Experiment 2, that result replicated when actors moved and when they were still. These results suggest participants may have picked up object height in relation to actor height. Implications about how people perceive other people’s maximum vertical one degree-of-freedom reach heights and how that might inform assistive robot design are discussed.


  • Grouping and Determining Perceived Severity of Cyber-Attack Consequences: Gaining Information Needed to Sonify Cyber-Attacks

    by Keith S. Jones, Natalie R. Lodinger, Benjamin P. Widlus, Akbar Siami Namin, Emily Maw, Miriam Armstrong

    Cyber-attacks are a continuing problem. These attacks are problematic for users who are visually impaired and cannot rely on visual cues to indicate a potential cyber-attack. Sonification is an alternative way to help users who are visually impaired detect potential cyber-attacks. Sonification provides information to users using non-speech sounds. Sonification could provide users who are visually impaired with information on potential cyber-attack consequences that could stem from their actions. However, there are two challenges with sonifying cyber-attack consequences. First, there are many potential cyber-attack consequences to sonify, and humans have a limited ability to remember associations between sonifications and their meanings. Second, cyber-attack warning messages are better trusted when they align the severity of the consequences with the user’s perceived severity. However, we do not know the perceived severity of individual consequences. Therefore, we need to reduce the number of consequences to sonify and to determine the perceived severity of these consequences. We had non-expert participants group cyber-attack consequences based on perceived similarity. Analyses revealed that participants’ groupings formed seven clusters. We then had non-expert participants rate the perceived severity of each cyber-attack consequence. Those ratings were used to determine the perceived severity of each cluster. These efforts resulted in a set of cyber-attack consequence clusters that (a) is small enough that users should be able to remember associations between sonifications and their meanings, and (b) can be sonified in a way that reflects users’ perceptions regarding the severity of the clustered cyber-attack consequences. As such, the results of these studies are critical steps towards creating effective sonifications that serve as cyber-security warning messages.


  • Open Access

    Markov Decision Process for Modeling Social Engineering Attacks and Finding Optimal Attack Strategies

    by Faranak Abri, Jianjun Zheng, Akbar Siami Namin, Keith S. Jones

    It is important to comprehend the attacker’s behavior and capacity in order to build a stronger fortress and thus be able to protect valuable assets more effectively. Prior to launching technical and physical attacks, an attacker may enter the reconnaissance stage and gather sensitive information. To collect such valuable data, one of the most effective approaches is through conducting social engineering attacks, borrowing techniques from deception theory. As a result, it is of utmost importance to understand when an attacker behaves truthfully and when the attacker opts to be deceitful. This paper models attacker’s states using Markov Decision Process (MDP) and studies the attacker’s decision for launching deception attacks in terms of cooperation and deception costs. The study is performed through Markov Decision Process (MDP) modeling, where the states of attackers are modeled along with the permissible actions that can be taken. We found that the optimal policy regarding being deceitful or truthful depends on the cost associated with deception and how much the attacker can afford to take the risk of launching deception attacks. More specifically, we observed that when the cost of cooperation is low (e.g., 10%), by taking MDP optimal policy, the attacker cooperates with the victim as much as possible in order to gain their trust; whereas, when the cost of cooperation is high (e.g., 50%), the attacker takes deceptive action earlier in order to minimize the cost of interactions while maximizing the impact of the attack. We report four case studies and simulations through which we demonstrate the trade-off between cooperative and deceptive actions in accordance with their costs to attackers.


  • Open Access

    A Comparative Analysis of Modeling and Predicting Perceived and Induced Emotions in Sonification

    by Faranak Abri, Luis Felipe Gutiérrez, Prerit Datta, David R. W. Sears, Akbar Siami Namin, Keith S. Jones

    Sonification is the utilization of sounds to convey information about data or events. There are two types of emotions associated with sounds: (1) “perceived” emotions, in which listeners recognize the emotions expressed by the sound, and (2) “induced” emotions, in which listeners feel emotions induced by the sound. Although listeners may widely agree on the perceived emotion for a given sound, they often do not agree about the induced emotion of a given sound, so it is difficult to model induced emotions. This paper describes the development of several machine and deep learning models that predict the perceived and induced emotions associated with certain sounds, and it analyzes and compares the accuracy of those predictions. The results revealed that models built for predicting perceived emotions are more accurate than ones built for predicting induced emotions. However, the gap in predictive power between such models can be narrowed substantially through the optimization of the machine and deep learning models. This research has several applications in automated configurations of hardware devices and their integration with software components in the context of the Internet of Things, for which security is of utmost importance.


  • How Do People Perceive Other People’s Affordances, and How Might That Help Us Design Robots That Can Do So?

    by Keith S. Jones, Nicholas A. Garcia

    A considerable amount of research has worked toward enabling robot caregivers to perform various tasks for individuals in need of assistance. However, little, if any, research has aimed to enable robot caregivers to determine when individuals need help performing tasks. One way to develop such robots is to start with what is already known about people determining whether other people can perform a task without help. Ecological Psychologists conceptualize that task in terms of people perceiving other people’s affordances. There is an extensive and growing literature concerning the perception of others’ affordances, which has provided many important insights. Hence, our long-term goal is to develop robot caregivers that perceive people’s affordances in ways that are similar to how people perceive others’ affordances, which will require a considerable amount of research. As a first step, we have carefully reviewed the Ecological Psychology literature concerning how people perceive other people’s affordances and discuss how such knowledge might inform the design of robot caregivers. In addition, we identify areas that, if further researched, would shed additional light on how to design robot caregivers that perceive people’s affordances, and move us toward a fuller understanding of how people perceive other people’s affordances.


  • Open Access

    Toward Explainable Users: Using NLP to Enable AI to Understand Users’ Perceptions of Cyber Attacks

    by Faranak Abri, Luis Felipe Gutiérrez, Chaitra T. Kulkarni, Akbar Siami Namin, Keith S. Jones

    To understand how end-users conceptualize consequences of cyber security attacks, we performed a card sorting study, a well-known technique in Cognitive Sciences, where participants were free to group the given consequences of chosen cyber attacks into as many categories as they wished using rationales they see fit. The results of the open card sorting study showed a large amount of inter-participant variation making the research team wonder how the consequences of security attacks were comprehended by the participants. As an exploration of whether it is possible to explain user’s mental model and behavior through Artificial Intelligence (AI) techniques, the research team compared the card sorting data with the outputs of a number of Natural Language Processing (NLP) techniques with the goal of understanding how participants perceived and interpreted the consequences of cyber attacks written in natural languages. The results of the NLP-based exploration methods revealed an interesting observation implying that participants had mostly employed checking individual keywords in each sentence to group cyber attack consequences together and less considered the semantics behind the description of consequences of cyber attacks. The results reported in this paper are seemingly useful and important for cyber attacks comprehension from user’s perspectives. To the best of our knowledge, this paper is the first introducing the use of AI techniques in explaining and modeling users’ behavior and their perceptions about a context. The novel idea introduced here is about explaining users using AI.