Syed Ibrahim Ghaznavi
Harnessing Multi-Agent LLMs for Complex Engineering Problem-Solving: A Framework for Senior Design Projects
Multi-Agent Large Language Models (LLMs) are gaining significant attention for their ability to harness collective intelligence in complex problem-solving, decision-making, and planning tasks. This aligns with the concept of the wisdom of crowds, where diverse agents contribute collectively to generating effective solutions, making it particularly suitable for educational settings. Senior design projects, also known as capstone or final year projects, are pivotal in engineering education as they integrate theoretical knowledge with practical application, fostering critical thinking, teamwork, and real-world problem-solving skills. In this paper, we explore the use of Multi-Agent LLMs in supporting these senior design projects undertaken by engineering students, which often involve multidisciplinary considerations and conflicting objectives, such as optimizing technical performance while addressing ethical, social, and environmental concerns. We propose a framework where distinct LLM agents represent different expert perspectives, such as problem formulation agents, system complexity agents, societal and ethical agents, or project managers, thus facilitating a holistic problem-solving approach. This implementation leverages standard multi-agent system (MAS) concepts such as coordination, cooperation, and negotiation, incorporating prompt engineering to develop diverse personas for each agent. These agents engage in rich, collaborative dialogues to simulate human engineering teams, guided by principles from swarm AI to efficiently balance individual contributions towards a unified solution. We adapt these techniques to create a collaboration structure for LLM agents, encouraging interdisciplinary reasoning and negotiation similar to real-world senior design projects. To assess the efficacy of this framework, we collected six proposals of engineeringand computer science of typical senior capstone projects and evaluated the performance of Multi-Agent and
single-agent LLMs using both custom-designed metrics developed in consultation with engineering faculty and
some widely used NLP-based metrics. These metrics cover technical quality, ethical considerations, social impact,
and feasibility, ensuring that our evaluation aligns with the educational objectives of engineering design. Our findings
suggest that Multi-Agent LLMs can provide a richer, more inclusive problem-solving environment compared to
single-agent systems, offering a promising tool for enhancing the educational experience of engineering and computer
science students by simulating the complexity and collaboration of realworld engineering and computer science practice.
By supporting senior design projects, this tool not only aids in achieving academic excellence but also prepares students
for the multifaceted challenges they will face in their professional engineering careers. [Paper archive link]
Toward Inclusive Educational AI: Auditing Frontier LLMs through a Multiplexity Lens
As large language models (LLMs) like GPT-4 and Llama 3 become integral to educational contexts, concerns are mounting over the cultural biases, power imbalances, and ethical limitations embedded within these technologies. Though generative AI tools aim to enhance learning experiences, they often reflect values rooted in Western, Educated, Industrialized, Rich, and Democratic (WEIRD) cultural paradigms, potentially sidelining diverse global perspectives. This paper proposes a framework to assess and mitigate cultural bias within LLMs through the lens of applied multiplexity. Multiplexity, inspired by Senturk et al. and rooted in Islamic and other wisdom traditions, emphasizes the coexistence of diverse cultural viewpoints, supporting a multi-layered epistemology that integrates both empirical sciences and normative values. Our analysis reveals that LLMs frequently exhibit cultural polarization, with biases appearing in both overt responses and subtle contextual cues. To address inherent biases and incorporate multiplexity in LLMs, we propose two strategies: Contextually-Implemented Multiplex LLMs, which embed multiplex principles directly into the system prompt, influencing LLM outputs at a foundational level and independent of individual prompts, and Multi-Agent System (MAS)-ImplementedMultiplex LLMs, where multiple LLM agents, each representing distinct cultural viewpoints, collaboratively generate
a balanced, synthesized response. Our findings demonstrate that as mitigation strategies evolve from contextual
improves, evidenced by a significant rise in the Perspectives Distribution Score (PDS) and a PDS Entropy increase
from 3.25% at baseline to 98% with the MAS-Implemented Multiplex LLMs. Sentiment analysis further shows a shift
towards positive sentiment across cultures, with the MAS-Implemented Multiplex LLMs achieving 0% negative
sentiment, underscoring enhanced cultural sensitivity. This pioneering study establishes a baseline for assessing and
fostering cultural inclusivity in educational AI, laying the groundwork for a globally pluralistic approach that respects
diverse cultural perspectives. We hope this work inspires further research toward creating AI technologies that serve
a truly inclusive and multicultural educational ecosystem. [Paper archive link]
Photorealistic avatars to enhance the efficacy of Self-attachment psychotherapy
We have designed, developed, and tested an Immersive virtual reality (VR) platform to practice the protocols of Self-attachment psychotherapy. We made use of customized photorealistic avatars for the implementation of both the high-end version (based on Facebook's Oculus) and the low-end version (based on Google's cardboard) of our platform. Under the Self-attachment therapeutic framework, the causes of mental disorders such as chronic anxiety and depression are traced back to the individual's insecure attachment with their primary caregiver during childhood and their subsequent problems in affect regulation. The conventional approach (without VR) to Self-attachment requires that the individual uses their childhood photographs to recall their childhood memories and then imagine that the child that they were is present with them. They thus establish a compassionate relationship with their childhood self and then, using love songs and dancing, create an affectional bond with them. Their adult self subsequently role plays a good parent and interacts with their imagined childhood self to perform various developmental and re-parenting activities to earn secure attachment. Our immersive virtual reality platform enables the users to interact with their customized 3D photorealistic childhood avatar in the VR. The platform takes advantage of the itSeez3D Avatar SDK for generating a customized photorealistic 3D avatar head from a 2D childhood image of the user. Furthermore, it allows modifications to the avatar body (height/ width) and clothing color. A study to compare the use of the avatar based approach (VR) to Self-attachment with the conventional photo based approach showed promising results; 85% of the participants reported that their photorealistic childhood avatar in VR was more relatable than their childhood photo. [Publication]Usability evaluation of an immersive virtual reality platform for Self-attachment psychotherapy
Virtual Reality (VR) is the state-of-the-art human-computer interface; it uses computer graphics to create a realistic-looking virtual world that the user can interact with in real-time. Recent advances in VR have shown promise in the pursuit of devising new techniques to combat mental disorder(s). Harnessing the power of VR, we have developed a customiseds immersive virtual reality platform to practise protocols of selfattachment psychotherapy. Consumer-level VR is a recent phenomenon; for wide scale adaptation of such platforms it is important that they are built taking into account usability engineering principles specific to virtual environments (VE). In this work, we share our experience applying systematic heuristic and formative evaluations to our VR platform to make it more usable. The participants who evaluated our platform were asked (via a follow-up questionnaire) to rate their level-of-immersion, learn-ability and overall usability of the platform. Insights from our usability evaluation could help developers build better and more usable psycho-therapeutic VR platforms in the future. [Publication]Immersive Virtual reality platform for Self-attachment psychotherapy
Psychotherapy is among the most effective techniques for combating mental health issues, and Virtual Reality is beginning to be explored as a way to enhance the efficacy of various psychotherapeutic treatments. In this work, we propose an Immersive Virtual Reality platform for Self-Attachment psychotherapy. Under the Self-Attachment therapeutic framework, the causes of disorders such as chronic anxiety and depression are traced back to the quality of the individual's attachment with their primary caregiver during childhood. Our proposed platform aims to assist the user in enhancing their capacities for self-regulation of emotion, by means of earning secure attachment through the experience of positive attachment interactions, missed in their childhood. In the virtual environment provided by the platform, the adult-self of the user learns to create and strengthen an affectional and supportive bond with the inner-child. It is hypothesised that by long term potentiation and neuroplasticity, the user gradually develops new neural pathways and matures into an effective secure attachment object for the inner-child, thereby enabling the self-regulation of emotions. [Publication]
Rescue Base Station - RBS
Rescue Base Station (RBS) is a drop-in, solar power compatible, open-source GSM communication system for scenarios where a large-scale calamity disrupts traditional modes of communication. The system operates using asynchronously connected autonomous nodes and gathers useful information from users, eventually synchronizing this data across the network using distributed network protocols. It connects people through conventional GSM services allowing calls, SMS and smart phone features when available. The networks also provides a series of services for use during a disaster, such as intelligent call routing, attribute based search on different characteristics (name, occupation and blood group), voice-mail services, SMS broadcast alerts, and emergency short-codes, through which a victim can contact available doctors, fire fighters, police and rescue workers. [Publication] [Presentation] [Code]