KFKianté Fernandez

Kianté Fernandez

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Kianté Fernandez has a total of 41 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Estimating the threat of AI-agent responding across online survey platforms

    by Grace Zhang, Robert Walatka, Stephanie Y. Chen, Alexander K. Moore, Oleg Urminsky, Kianté Fernandez, Andrea Low, Jonathan E. Bogard, Craig R. Fox

    Recent research and advances in LLMs have led to widespread concern that AI agents could pose as human online survey-takers. However, it remains unclear how prevalent AI agents are on survey research platforms and how to effectively detect AI agents. We validated a series of AI detection tests that effectively separated verified-human participants from both prompted commercial and purpose-built AI agents. Using these tests, we collected surveys on seven online platforms and find wide variation in rates of respondents failing AI tests, ranging from 6 to 41% across platforms (compared to a 2.4% false-positive rate for in-person human participants). We further demonstrate that undetected AI agents can impact the results of online surveys. Our findings suggest that while there is an urgent need for AI detection tests and consistent, systematic monitoring of data quality on online platforms, some platforms currently appear to provide data with a low rate of AI agents.


  • Open Access

    A database of subjective evaluation ratings for decision-making research

    by Kianté Fernandez, Sumedha Goyal, Ian Krajbich

    Subjective value ratings guide everyday choices, and value-based decision-making research collects them in many experiments. Yet the field has no common method for eliciting them. Studies use different response scales, instructions, and stimulus sets, and their ratings are rarely shared in a form that allows them to be used together. We present the Liking Initiative, a curated database of 759,399 individual ratings from 38 studies (59 datasets) covering 2,217 items, rated by 5,377 participants summed across datasets. Each rating is provided along with dataset-level metadata documenting the response scale, the question asked, and the source publication. The database is available through a web application, versioned releases on Zenodo, and R and Python packages. Because many items were rated in more than one dataset, item-level ratings can be compared across samples, response scales, and laboratories.

  • Open Access

    Linking Traits and Behavior Through Computational Modeling of Neuroethical Decisions

    by Brian Kim, Kianté Fernandez, John Medaglia

    Individual differences in psychology are often measured through self-report questionnaires or behavioral performance, yet these approaches capture different constructs and rarely align. Computational modeling provides a framework for connecting both measures, as it extracts a process-level description of cognition from observable behavior in the form of latent variables. These latent variables represent repeated patterns of decision-making that should relate to psychological traits. We applied this approach to neuroethical decision-making, where evidence shows that both behavioral performance and self-reported psychological traits are related to how individuals weigh uncertain risks against potential benefits of emerging brain technologies. Participants completed a neuromodulation task in which risk and benefit probabilities were systematically varied. To capture how decisions unfolded, we used a temporally sensitive diffusion decision model that estimates when and how risks and benefits are evaluated. These model parameters described general trends in task behavior: individuals considered risks earlier and weighed risks more highly than benefits. Additionally, model parameters described how individuals modified their behavior: individuals who considered benefits first according to the model tended to endorse neuromodulation, but shifted towards rejection when risks were evaluated earlier. Importantly, self-reported psychological traits related to these computational parameters. Cognitive factors such as risk propensity and numeracy related to risk sensitivity and evaluation timing, while prosocial traits were also related to evaluation timing and benefit sensitivity. By contrast, Religion & Spirituality and Moral Commitment measures showed lower than expected associations across parameters. By capturing process-level mechanisms, computational modeling served as a bridge between self-reported traits and observed behavior, showing how stable individual differences shape the dynamics of decision-making in neuroethical contexts.

  • Open Access

    The first-passage time density for the diffusion model with variable drift and variable starting point

    by KiantÉ Fernandez

    The Ratcliff diffusion model is fitted with across-trial variability in drift rate, starting point, and non-decision time. The drift integral of the first-passage time density has been evaluated in closed form at fixed starting point, and the starting-point and non- decision-time integrals at fixed drift, but the drift and starting-point integrals have not been evaluated together. In particular, existing software implementations integrate at least one of these two numerically. We show that in the small-time series representation the starting-point integral of the drift-integrated density is an integral of a linear function against a normal density, and we give the density with normally distributed drift and uniformly distributed starting point in closed form together with a large-time counterpart in terms of the complex error function. The six-parameter density then requires no numerical integration at all, and the seven- parameter likelihood a single one-dimensional quadrature. The derivatives of the density with respect to all of its parameters follow in closed form, and the likelihood gradient reuses that one quadrature. The result is verified symbolically and against three reference implementations, and is machine-checked in Lean.

  • masc: Simulate the Multi-Attribute Search and Choice (MASC) Model

    by Kiante Fernandez

    Simulates the Multi-Attribute Search and Choice (MASC) model of Gluth, Deakin and Rieskamp (2026) for multi-attribute decision-making, including sequential information search, Bayesian belief updating, and choice. Beliefs may be treated as univariate (independent attributes), or multivariate over correlated attributes ('MASC-C'), in which observing one attribute updates beliefs about correlated attributes via a Kalman filter.

  • Open Access

    Defending Against AI Agents with Webcam Eye Tracking

    by Kianté Fernandez, Xiaozhi Yang, Stephanie Marie Smith, Ian Krajbich

    Autonomous AI agents, now embedded in everyday browsers, can complete online surveys and experiments with the click of a button, threatening the integrity of behavioral research that depends on online samples. Existing defenses, such as CAPTCHAs and attention checks, are easily bypassed. Webcam eye tracking offers a different defense: proof by presence. Unlike behavioral signals that AI agents increasingly mimic, gaze relies on physiological signals that are not feasible to counterfeit. We introduce a compendium of open-source tools that integrate webcam eye tracking into the platforms responsible for most online behavioral research: Qualtrics for surveys; jsPsych and PsychoPy for cognitive experiments; and oTree for experimental economics. We also provide guidance for researchers to screen out AI agents without penalizing human participants.

  • Open Access

    The threat of synthetic respondents extends to clinical mental health screening

    by Kianté Fernandez, Laura Berner, Blair R K Shevlin

    Online platforms have enabled clinical mental health research at unprecedented scale, yet large language models (LLMs) now threaten the integrity of remotely collected data. Recent work has demonstrated that LLM-based synthetic respondents can complete general surveys with human-like plausibility, but the vulnerability of validated clinical psychiatric instruments remains unexamined. Here, we show that a commercially available LLM, given only brief diagnostic persona descriptions, produces clinically differentiated responses across seven standardized psychiatric screening instruments spanning mood, anxiety, obsessive-compulsive, trauma, psychotic, neurodegenerative, and eating disorders. In a simulation of 2106 synthetic personas, diagnosis-congruent personas generated scores exceeding clinical cutoffs on five of seven instruments, with scores scaling monotonically with assigned severity. This mimicry required no specialized configuration. Coherent symptom endorsement can no longer serve as a proxy for authentic participation, urgently necessitating standardized verification methods for clinical online data.

  • Open Access

    Estimating the threat of AI-agent responding across online survey platforms

    by Stephanie Chen, Oleg Urminsky, Grace Zhang, Robert Walatka, Kianté Fernandez, Andrea Low, Jonathan Bogard, Craig R. Fox

    Recent research and advances in LLMs have led to widespread concern that AI agents could pose as human online survey-takers. However, it remains unclear how prevalent AI agents are on these platforms and how to effectively detect AI agents. We validated a series of AI detection tests that effectively separated verified-human participants from three AI agents (designed using various prompts). Using these tests, we collected surveys on seven online platforms and find high variance in rates of participants failing AI tests, ranging from 6% to 41% across platforms (compared to a 2.4% in-person human false-positive rate). We demonstrate that undetected AI agents can impact the results of online surveys. Our findings suggest that while there is an urgent need for AI detection tests and consistent, systematic monitoring of data quality on online platforms, currently some platforms seem to provide data with a low rate of AI agents.


  • Open Access

    A little goes a long way: Fitting one-shot decisions with cognitive models

    by Blair R K Shevlin, Kianté Fernandez, Roger Ratcliff, Ian Krajbich

    Cognitive models provide profound insights into the psychological processes underlying behavior. However, a significant limitation has constrained their application: the reliance on extensive, repeated-trial data from each participant. The data-hungry nature of cognitive models has largely precluded their application to infrequent but consequential one-shot decisions common in economic, social, and clinical contexts, and has excluded populations unable to complete lengthy experiments. Here, we address this methodological constraint by proposing a conceptual shift: instead of requiring many trials from a few individuals, we leverage few trials from many individuals. By treating between-subject variability as a source of information, we demonstrate that cognitive models can be successfully fit to one-shot data. Through a series of simulations, we first establish that we can recover known parameter values from single-trial data. We then validate the approach empirically by showing that with only a single trial per participant it can replicate canonical findings; namely, the speed-accuracy trade-off, the influence of food quality and expectations on choice, and the task-specificity of linear vs. non-linear numeric representations. This work overcomes a major limitation to the widespread application of cognitive models, opening new frontiers for understanding the cognitive mechanisms of real-world choices.

  • Open Access

    SequentialSamplingModels.jl: Simulating and Evaluating Cognitive Models of Response Times in Julia

    by Kianté Fernandez, Dominique Makowski, Christopher Fisher