Organizational Unit:

School of Psychology

Permanent Link

https://hdl.handle.net/1853/70756

Parent Organization

Organizational Unit

College of Sciences

Includes Organization(s)

Organizational Unit

Work Science Center

ArchiveSpace Name Record

https://finding-aids.library.gatech.edu/agents/corporate_entities/1133

Full item page

Publication Search Results

Now showing 1 - 10 of 22

Measure and Manage Trust in Human-AI Conversations

(Georgia Institute of Technology, 2023-10-12) Li, Mengyao

Artificial Intelligence (AI), with its increasing capability and connectivity, extends beyond limited and well-defined contexts and is integrated into broader societal domains. AI algorithms now steer autonomous vehicle fleets, shape political beliefs through news filtering, and oversee resource allocation and labor. Establishing trust between humans and their AI counterparts becomes important to facilitate effective cooperation. Trust profoundly influences how individuals use, communicate with, and collaborate alongside AI systems. Thus, trust measurement and management within human-AI cooperation are indispensable for ensuring safety, efficiency, and overall success. This talk focuses on trust in human-AI interactions, addressing three primary questions: (1) How can we measure people’s trust in human-AI conversations? (2) How does trust change over time within human-AI conversations? (3) How can we effectively manage instances of overtrust or undertrust through conversational cues to enhance human-AI cooperation? This talk highlights critical advancements in measurement of trust dynamics in human-AI cooperation, promising to influence the future of AI integration into broader societal domains.
Deep Thinking about Deepfake Videos: Understanding and Bolstering Humans’ Ability to Detect Deepfakes

(Georgia Institute of Technology, 2023-03-16) Tidler, Zachary

“Deepfakes” are videos in which the (usually human) subject of a video has been digitally altered to appear to do or say something that they never actually did or said. Sometimes these manipulations produce innocuous novelties (e.g., testing what it would look like if Will Smith had been cast as “Neo” in the film The Matrix), but far more dangerous use cases have been observed (e.g., producing fake footage of Ukrainian President, Volodymyr Zelenskyy, in which he instructs Ukrainian military forces to surrender on the battlefield). Generating the knowledge and tools necessary to defend against potential harms these videos could impose is likely to rely on contributions from a broad coalition of disciplines, many of which are represented in the GVU. In this week’s Brown Bag presentation, we will offer some real-time demonstrations of deepfake technology and present findings from our work that has largely focused on investigating the psychological factors influencing deepfake detection.
Modeling Developmental Processes Using Accelerated Cohort-Sequential Data

( 2021-11-10) Ferrer, Emilio

Studying the time-related course of psychological processes is a challenging endeavor, particularly over long developmental periods. Accelerated longitudinal designs (ALD) allow capturing such periods with a limited number of assessments in a much shorter time framework. In ALDs, participants from different age cohorts are measured repeatedly but the measures provided by each participant cover only a fraction of the study period. It is then assumed that the common trajectory can be studied by aggregating the information provided by the different converging cohorts. In this presentation, I report results from recent studies examining the performance of discrete- and continuous-time latent change score (LCS) models for recovering the trajectories of a developmental process from data obtained through different ALDs. These results support the effectiveness of LCS models to study developmental processes using data from ALDs under various conditions of sampling. When all cohorts are drawn from the same population, both types of models are able to recover the parameters defining the underlying developmental process. However, discrete-time models yield estimates with bias when time lags between observations are not constant. When cohorts are not from the same population and lack convergence, both discrete- and continuous-time models show bias in some dynamic parameters.
How to Kill Zombie Ideas: Why do people tenaciously believe myths about the relationship between people & technology?

( 2021-10-27) Woods, David

Zombie ideas plague much of the discussions on deploying AI and other autonomous machine capabilities into fields of human activity. People consistently mis-envision the impact of deploying these technologies by a wide mark. Because these oversimplified and erroneous ideas about AI and autonomy reappear and persist even after repeated empirical and technical debunking, they are zombies. This gives rise to the query: how can we kill off zombies ideas, in this case, about AI and Autonomy to be deployed into Systems that Serve Human Purposes. As this question recurs over decades, and because the zombies are prevalent among technolologists, social scientists, and management ranks, there must be deep psychological roots to these misconceptions. Ultimately, these enduring misconceptions, when recognized as zombies, represent a path to see deeply into the ultimate nature of our biological universe. To prepare ahead, try to list as many candidate myths about the relationship between people & technology as you can generate yourself or together with others. Second, think about the kind of reasoning fallacies that contribute to zombie ideas. Third can you deconstruct 3 popular oxymorons that appear to be essential yet are really zombies in disguise: ethical algorithms, explainable AI, assured autonomy. Warning: Pointing out zombies/myths is a highly dangerous activity. Just listening to, much less participating in, deconstructing these myths can lead to disorientation, anxiety, confusion & the urge to defend zombies.
General Intelligence Explained (Away)

( 2021-03-18) Conway, Andrew R. A.

For more than a century, the standard view in the field of human intelligence has been that there is a “general intelligence” that permeates all human cognitive activity. This general cognitive ability is supposed to explain the positive manifold, the finding that intelligence tests with different content all correlate. However, this interpretation does not sit well with findings from cognitive psychology and neuroscience that point to the domain-specific modular fractionation of cognition. In my research talk I will present an alternative interpretation - process overlap theory - a new theoretical framework for the study of individual differences in cognitive ability (Conway & Kovacs, 2013; 2015; Kovacs & Conway, 2016; 2019). The theory assumes that most forms of complex cognition, and IQ test items, require a number of domain-general as well as domain-specific processes. Domain-general processes involved in executive attention are central to test performance. That is, they are activated by a large number of test items, alongside with domain-specific processes tapped by specific types of tests only. Such an overlap of executive processes explains the positive manifold as well as the hierarchical structure of cognitive abilities and rejects the notion of a general mental ability. As a consequence of the theory, IQ is redefined as an emergent formative construct rather than a reflective latent trait. This implies that IQ should be interpreted as an index of specific cognitive abilities rather than the reflection of an underlying general cognitive ability. The consequences of this new approach will be discussed, including a focus on specific abilities rather than on global measures of cognitive performance.
Perceptual-Cognitive Expertise

( 2021-02-18) Willliams, Andrew Mark
Resisting the Knowledge Dementors: The Truth about “Post-Truth”

( 2021-02-04) Lewandowsky, Stephan

We are said to live in a “post-truth” era in which “fake news” has replaced real information, denial has compromised science, and the ontology of knowledge and truth has taken on a relativist element. I argue that to defend evidence-based reasoning and knowledge against those attacks, we must understand the strategies by which the post-truth world is driven forward. I depart from the premise that the post-truth era did not arise spontaneously but is the result of a highly effective political movement that deploys a large number of rhetorical strategies. I focus on three strategies: The deployment of conspiracy theories, the use of “micro-targeting” and “bots” online, and agenda-setting by attentional diversion. I present evidence for the existence of each strategy and its impact, and how it might be countered.
Learning distributed representations in the human brain

( 2020-11-19) Schapiro, Anna

The remarkable success of neural network models in machine learning has relied on the use of distributed representations — activity patterns that overlap across related inputs. Under what conditions does the brain also rely on distributed representations for learning? There are benefits and costs to this form of representation: it allows rapid, efficient learning and generalization, but is highly susceptible to interference. We recently developed a neural network model of the hippocampus that proposes that one subregion (CA1) may employ this form of representation, complementing known pattern-separated representations in other subregions. This provides an exciting domain to test ideas about learning with distributed representations, as the hippocampus learns much more quickly than the neocortical areas that have often been proposed to contain these representations. I will present modeling and empirical work that provide support for the idea that parts of the hippocampus do indeed learn using distributed representations. I will also present ideas about how hippocampal and neocortical areas may interact during sleep to further transform these representations over time.
Temporal oscillations in preference strength provide evidence for a quantum-Markov open system model of preference evolution

( 2020-11-05) Busemeyer, Jerome R.

We examined how preferences evolve across time in two new experiments, one using choices between restaurants and a second using choices between gambles. In both we observed that mean preference strength systematically oscillated over time and found that eliciting a choice early in time strongly affected the pattern of preference oscillation later in time. Preferences following choices oscillated between being stronger than those without prior choice and being weaker than those without choice. Markov processes, such as random walk models, have been successfully used by cognitive and neural scientists to model human choice behavior and decision time for over 50 years. Recently, quantum walk models have been introduced as an alternative way to model the dynamics of human choice and confidence across time. Our new findings point to the need for both types of processes, and what are called “open system” models provide a way to incorporate them both into a single process. The open system model incorporates two sources of uncertainty: epistemic uncertainty about what preference state a decision maker has at a particular point in time; and ontic uncertainty about what decision or judgment will be observed when a person has some preference state. Representing these two sources of uncertainty allows the model to account for the oscillations in preference as well as the effect of choice on preference formation.
It's the end of testing as we know it... and I feel fine

( 2020-10-29) Kingston, Neal

Neal Kingston will not sing about where testing needs to go. That's great, testing needs an earthquake; E C D, and D C M, and Embretson is not afraid; Eye of a hurricane, Markov chains must churn; Testing serves its own needs, must serve learning needs; Instruct embedded tests, grunt, more strength; Learning ladder starts to clatter with a fearsome trajectory; Valid’s looking pallid, represent with stealth games; Higher standards are for hire on a cheating site; Formative assessment in a hurry with the Feds breathing down your neck; State by state, teachers baffled, trumped, tethered, cropped; Look at those low scores, fine, then; Uh¬oh, over-tested, population, common core; But it'll do, test yourself, best yourself; Testing purpose ask why, students and the teachers cry; Tell me your scores are valid and your consequences are right, right; You’re estimated, validated, pub. fight, pub. right; Learning map it, psyched; It's the end of testing as we know it; It's the end of testing as we know it; It's the end of testing as we know it, and I feel fine.