Dr. Othalia Larue, Parallax director of cognitive AI, was born in France, where she studied and received her master's degree in computer science from Epitech. Following her master’s, Dr. Larue continued her studies at the University of Quebec in Montreal, Canada, where she researched cognitive architectures. Today, she is based in Dayton, Ohio and continues her research focused on cognitive architectures at Parallax Advanced Research.
We sat down with Dr. Larue to learn about what her team is working on and the exciting new ventures in cognitive AI.
Q: How do you define Cognitive AI, and what are the areas where it's making the most significant impact today?
Dr. Larue: Cognitive AI takes inspiration from the way humans perceive, process information and reason to design AI solutions. As such, it can address a wide range of complex tasks that would require human-like cognitive abilities. I define my job as both: (1) modeling cognitive processes and capturing individual differences in the way humans process information and reason, and (2) leveraging those models to create algorithms that can be used in the design of AI solutions.
Q: Parallax Advanced Research is known for its cutting-edge work in AI and cognitive technologies. Could you share some recent projects or innovations that your team has been working on?
Dr. Larue:
My expertise is in cognitive architectures, computational instantiations of the mind. They act both as unified theories of cognition (I.e. theories that, instead of focusing on one specific effect of cognition, explain cognitive phenomenon as they emerge from the interaction of multiple processes) and their software implementation. I believe these architectures will keep driving progress in the field of cognitive modeling where this type of integrative approach is needed. There are multiple cognitive architectures developed by different research groups, each having different strengths. I wrote a book chapter about this; see Larue, O., Bourdon, J. N., Legault, M., & Poirier, P. (2022). Mental Architecture—Computational Models of Mind. In Mind, Cognition, and Neuroscience (pp. 164-182). Routledge.
I developed my own cognitive architecture during my PhD and have been given the opportunity recently to work on a hybrid architecture with Dr. Alex Hough from the Air Force Research Laboratory, combining the companion architecture’s analogies capacities that were developed by Dr. Ken Forbus’s group at Northwestern University and the ACT-R cognitive architecture’s memory capacities that were developed Dr John Anderson and Dr Christian Lebiere at Carnegie Mellon University. The resulting hybrid cognitive architecture provided augmented problem-solving capacities.
I am also very excited about an upcoming project with the U.S. Air Force’s Autonomy Capabilities Team, where we will be working with the LEABRA cognitive architecture, developed by Dr Randall O’Reilly from the University of California - Davis. This will allow us to model finer grained cognitive processes, such as perception, and I really look forward to transitioning some of the algorithms we develop to AI solutions.
The Parallax AI and Autonomy Team has been working on the Defense Advanced Research Projects Agency In The Moment (ITM) project, focused on human-aligned decision-making in small-unit triage in austere environments and mass casualty care. In these environments, there are no perfect solutions and even experts do not agree on a course of action. Parallax's solution, called “Trustworthy Algorithmic Delegate (TAD),” under development, is an innovative explainable case-based reasoning (ECBR) system. We are collaborating with Professor Rosina Weber from Drexel University and her team, as well as Knexus Research Corporation. We will present our progress on TAD at the upcoming 2024 AAAI spring symposium on human-like learning. Learn more about this research here: Rauch C., Addison U., Floyd M., Prateek G., Karneeb J., Kulhanek R., Larue O., Ménager D., Mainali M., Molineaux M., Pease A., Sen A., Turner JT. and Weber R. Algorithmic Decision-Making in Difficult Scenarios. To achieve trust, measured as the expert’s willingness to delegate decision-making authority, TAD will attempt to align its decisions to a trusted human decision-maker. Specifically in TAD, Parallax Research Intern, Ursula Addison and I are working on designing heuristic rules inspired by fast, frugal, and naturalistic decision-making heuristics, which are observed human decision-making strategies that have been employed in several fields (such as medical) and reduce the complexity of real environments by employing a minimum of time, knowledge, and computations.
Q: How does your team address ethical concerns in the development of cognitive AI technologies?
Dr. Larue: On the ITM project, performers are working in collaboration with the University of Maryland Applied Research Laboratory for intelligence and security and the Institute for Defense Analyses for policy/practice integration and outreach, including ethical, legal, and societal implications experts throughout the research process.
For the Autonomy Capabilities Team, an AI special operations organization, whose mission is to operationalize AI at scale for the U.S. Air Force, in collaboration with Dr. Matt Molineaux, director of the Parallax’s AI and Autonomy Program, and Parallax Research Intern Ursula Addison, we built a proof of concept of an agent designed to assist an operator’s ethical decision-making in a plant protection toy domain. The system’s morality is a hybrid of consequentialism and deontology, with a degree of configurability based on the norms and values of its controlling organization. Agents created with this system would be able to inform human operators about the ethical implications of a goal. Specifically, by incorporating both automatic decisions aligned with the value of a stakeholder, as well as the more deliberative type of decisions, and providing the operators information about elements and confidence levels that lead to engaging in more deliberation, this agent would support the operator to produce more informed ethical decisions aligning with its value. This proof of concept was published at the RaD-AI23: Rebellion and disobedience in AI workshop.
Q: What trends do you see on the horizon for cognitive AI, and how do you envision Parallax Advanced Research contributing to these developments?
Dr. Larue: Specifically for Parallax, human-machine teaming is an aspect that requires cognitive modeling of the user to better tailor the solutions you’re developing to individual differences in cognitive processing (e.g. What is the current situational awareness of an operator? Does he have the cognitive workload to take on a new task? etc.). Trust in automation is a research area in which there are currently a lot of progress. We need to design tools that consider individual differences in trusting behavior (I.e. initial propensity to trust, operator experience, trust building and restoration dynamics). There needs to be a balance between not only building and maintaining trust but also adequately calibrating trust, so people aren’t blindly trusting the machine when performance isn’t there.
We built a prototype trust in automation algorithm that was first supported by Parallax’s internal Independent Research & Development (IR&D) initiative and now the U.S. Air Force’s Autonomy Capabilities Team, which considers the impact of positive and negative events in interaction with the automation (i.e., more specifically the rate at which they are memorized and forgotten) and how it impacts future delegation decisions.
Long-term plans for the ITM project entail our working on tailoring explanations to specific decision-makers with different cognitive styles. This includes adequately presenting which information were part of the decision by presenting, for example, the intermediary results of search heuristics they might adopt to parse the information efficiently.
Q: What advice do you have for aspiring AI researchers and innovators who want to make a significant impact in the field?
Dr. Larue: The advice I would give is to keep learning and remain open to new information. You must be ready for experiences or jobs that change the trajectory of your career, so don’t be too stubborn with what it is you think you want to do. Let yourself be surprised by interesting new opportunities. I would also advise you to remain "scientifically curious”. Attend conferences, read papers that are not only from your field but are broader and inform you of other’s research in the ecosystem. There are always interesting connections to make with other fields and researchers, learn about other methods that then can, in turn, inform your own research. Interacting with both AI scientists and psychologists has been a big contribution to my research, it keeps me creative. Don’t work in isolation.
Q: Lastly, could you highlight a particularly memorable or rewarding moment from your career in the field of Cognitive AI?
Dr. Larue: I have recently been given the opportunity to continue working on a few original ideas. I have been working with Dr Alex Hough from the Air Force Research Laboratory on modeling the Continued Influence Effect (I.e. misinformation can have a lasting effect even after retraction). Cognitive modeling has been scarcely used in establishing a psychology of misinformation, and I believe this could drive progress. This work leverages modeling work that I have been passionate about since my PhD. It is very rewarding to see some of those ideas concretize and be applied to such an interesting and current problem.
###
About Parallax Advanced Research
Parallax Advanced Research is a 501(c)(3) private nonprofit research institute that tackles global challenges through strategic partnerships with government, industry, and academia. It accelerates innovation, addresses critical global issues, and develops groundbreaking ideas with its partners. With offices in Ohio and Virginia, Parallax aims to deliver new solutions and speed them to market. In 2023, Parallax and the Ohio Aerospace Institute formed a collaborative affiliation to drive innovation and technological advancements in Ohio and for the Nation. The Ohio Aerospace Institute plays a pivotal role in advancing the aerospace industry in Ohio and the nation by fostering collaborations between universities, aerospace industries, and government organizations and manages aerospace research, education, and workforce development projects. More information about both organizations can be found at https://parallaxresearch.org/ and https://oai.org/.