Human Cognition Research Articles

ViSTec: Video Modeling for Sports Technique Recognition and Tactical Analysis

The immense popularity of racket sports has fueled substantial demand in tactical analysis with broadcast videos. However, existing manual methods require laborious annotation, and recent attempts leveraging video perception models are limited to low-level annotations like ball trajectories, overlooking tactics that necessitate an understanding of stroke techniques. State-of-the-art action segmentation models also struggle with technique recognition due to frequent occlusions and motion-induced blurring in racket sports videos. To address these challenges, We propose ViSTec, a Video-based Sports Technique recognition model inspired by human cognition that synergizes sparse visual data with rich contextual insights. Our approach integrates a graph to explicitly model strategic knowledge in stroke sequences and enhance technique recognition with contextual inductive bias. A two-stage action perception model is jointly trained to align with the contextual knowledge in the graph. Experiments demonstrate that our method outperforms existing models by a significant margin. Case studies with experts from the Chinese national table tennis team validate our model's capacity to automate analysis for technical actions and tactical strategies. More details are available at: https://ViSTec2024.github.io/.

Automatic Scoring of Metaphor Creativity with Large Language Models

ABSTRACT Metaphor is crucial in human cognition and creativity, facilitating abstract thinking, analogical reasoning, and idea generation. Typically, human raters manually score the originality of responses to creative thinking tasks – a laborious and error-prone process. Previous research sought to remedy these risks by scoring creativity tasks automatically using semantic distance and large language models (LLMs). Here, we extend research on automatic creativity scoring to metaphor generation – the ability to creatively describe episodes and concepts using nonliteral language. Metaphor is arguably more abstract and naturalistic than prior targets of automated creativity assessment. We collected 4,589 responses from 1,546 participants to various metaphor prompts and corresponding human creativity ratings. We fine-tuned two open-source LLMs (RoBERTa and GPT-2) – effectively “teaching” them to score metaphors like humans – before testing their ability to accurately assess the creativity of new metaphors. Results showed both models reliably predicted new human creativity ratings (RoBERTa r = .72, GPT-2 r = .70), significantly more strongly than semantic distance (r = .42). Importantly, the fine-tuned models generalized accurately to metaphor prompts they had not been trained on (RoBERTa r = .68, GPT-2 r = .63). We provide open access to the fine-tuned models, allowing researchers to assess metaphor creativity in a reproducible and timely manner.

The Fairness Fair: Bringing Human Perception into Collective Decision-Making

Fairness is one of the most desirable societal principles in collective decision-making. It has been extensively studied in the past decades for its axiomatic properties and has received substantial attention from the multiagent systems community in recent years for its theoretical and computational aspects in algorithmic decision-making. However, these studies are often not sufficiently rich to capture the intricacies of human perception of fairness in the ambivalent nature of the real-world problems. We argue that not only fair solutions should be deemed desirable by social planners (designers), but they should be governed by human and societal cognition, consider perceived outcomes based on human judgement, and be verifiable. We discuss how achieving this goal requires a broad transdisciplinary approach ranging from computing and AI to behavioral economics and human-AI interaction. In doing so, we identify shortcomings and long-term challenges of the current literature of fair division, describe recent efforts in addressing them, and more importantly, highlight a series of open research directions.

Dialogues Are Not Just Text: Modeling Cognition for Dialogue Coherence Evaluation

The generation of logically coherent dialogues by humans relies on underlying cognitive abilities. Based on this, we redefine the dialogue coherence evaluation process, combining cognitive judgment with the basic text to achieve a more human-like evaluation. We propose a novel dialogue evaluation framework based on Dialogue Cognition Graph (DCGEval) to implement the fusion by in-depth interaction between cognition modeling and text modeling. The proposed Abstract Meaning Representation (AMR) based graph structure called DCG aims to uniformly model four dialogue cognitive abilities. Specifically, core-semantic cognition is modeled by converting the utterance into an AMR graph, which can extract essential semantic information without redundancy. The temporal and role cognition are modeled by establishing logical relationships among the different AMR graphs. Finally, the commonsense knowledge from ConceptNet is fused to express commonsense cognition. Experiments demonstrate the necessity of modeling human cognition for dialogue evaluation, and our DCGEval presents stronger correlations with human judgments compared to other state-of-the-art evaluation metrics.

Investigating the Relationship between Noise Exposure and Human Cognitive Performance: Attention, Stress, and Mental Workload Based on EEG Signals Using Power Spectrum Density

A pervasive environmental stressor is one that damages mental and physical health as well as cognitive abilities by producing noise at a specific frequency and level. Current noise pollution levels pose a significant threat to public health, potentially leading to impaired cognitive function, increased stress, and other negative health consequences. This study aims to investigate the relationship between noise exposure and human cognitive abilities using a comprehensive analysis of power spectrum density (PSD) derived from EEG signals. Twenty-four participants completed the experiment to identify the effect of exposure to different noise levels (55 dB, 65 dB, 70 dB, 75 dB, 80 dB, and 85 dB) and two types of continuous and intermittent noise. The Stroop Color–Word Test and the Emotive Epoch EEG are cognitive task instruments used during experiments. Behavioral performance (accuracy and response time) and power spectrum electroencephalographic density were collected and analyzed. The methodology involved collecting EEG data from participants exposed to controlled noise stimuli and a subsequent PSD analysis to uncover frequency-specific patterns associated with cognitive processes. Attention levels were measured by examining beta wave activity, while stress responses were evaluated through an alpha wave analysis. Additionally, mental workload was assessed by considering the overall distribution of PSD through the theta-to-alpha ratio. The results revealed a significant relationship between the exposure to noise types and levels and human cognitive ability. The analysis of the power spectrum density on the cognitive aspects of attention and stress yielded results indicating that participants were in the best attention condition and in a relaxed or unstressed state when exposed to noise levels of 65 dB in both continuous and intermittent noise types. For the mental workload aspect, participants exposed to both continuous and intermittent noise types at a noise level of 70 dB began to indicate the presence of mental workload. These findings supported the importance of considering the impact of environmental noise on human cognitive well-being and demonstrated the potential of EEG monitoring as an objective tool for assessing the impact of noise on cognitive performance.

Understanding the operators’ cloud change errors based on cognitive abilities and personality traits: An investigation integrated with quantitative and qualitative methods

While cloud services make industrial data convenient, they also expose it to cloud incidents. Operators' error in cloud change activities is a leading factor for cloud incidents, which have received relatively less attention in cloud security research. This study conducted a two-stage research process using an integrated approach to explore the stable individual factors related to cloud change errors. First, in the qualitative research, content analysis based on interviews and historical documents was conducted to extract the operator's cognitive abilities and personality traits and develop hypotheses. Five cognitive abilities and six personality traits were extracted. Second, quantitative research based on an experiment was conducted to test relationships between operators' different types of cloud change errors and 1) cognitive ability and 2) personality traits, respectively. Results of error type comparisons suggested that operators generated more uncorrected errors than corrected errors and more operational errors than omission errors in cloud change activities. The multivariate Poisson regression analysis suggested that cognitive abilities of sustained attention, divided attention, and long-term memory negatively predicted the number of operators' total errors, uncorrected errors, and operational errors. Regarding personality traits, with the increase in resilience capacity and carefulness and the decrease in self-esteem, the number of different types of errors reduced, except for omission errors. Working memory and risk-taking propensity were also significant predictors of the number of uncorrected errors with negative and positive coefficients, respectively. Logical reasoning, emotional stability, and sense of responsibility were not observed as predictors of cloud change errors. The present findings have several implications for the industry and cloud providers to enhance industrial cloud data security regarding human cognitive abilities and personality traits.

The neural basis of naturalistic semantic and social cognition

Decoding social environments and engaging meaningfully with other people are critical aspects of human cognition. Multiple cognitive systems, including social and semantic cognition, work alongside each other to support these processes. This study investigated shared processing between social and semantic systems using neuroimaging data collected during movie-viewing, which captures the multimodal environment in which social knowledge is exchanged. Semantic and social content from movie events (event-level) and movie transcripts (word-level) were used in parametric modulation analyses to test (1) the degree to which semantic and social information is processed within each respective network and (2) engagement of the same cross-network regions or the same domain-general hub located within the semantic network during semantic and social processing. Semantic word and event-level content engaged the same fronto-temporo-parietal network and a portion of the semantic hub in the anterior temporal lobe (ATL). Social word and event-level content engaged the supplementary motor area and right angular gyrus within the social network, but only social words engaged the domain-general semantic hub in left ATL. There was evidence of shared processing between the social and semantic systems in the dorsolateral portion of right ATL which was engaged by word and event-level semantic and social content. Overlap between the semantic and social word and event results was highly variable within and across participants, with the most consistent loci of overlap occurring in left inferior frontal, bilateral precentral and supramarginal gyri for social and semantic words and in bilateral superior temporal gyrus extending from ATL posteriorly into supramarginal gyri for social and semantic events. These results indicate a complex pattern of shared and distinct regions for social and semantic cognition during naturalistic processing.Protocol registrationThe stage 1 protocol for this Registered Report was accepted in principle on October 11, 2022. The protocol, as accepted by the journal, can be found at: https://doi.org/10.17605/OSF.IO/ACWQY.

Quantum-Mechanical Modelling of Asymmetric Opinion Polarisation in Social Networks

We propose a quantum-mechanical model that represents a human system of beliefs as the quantised energy levels of a physical system. This model represents a novel perspective on opinion dynamics, recreating a broad range of experimental and real-world data that exhibit an asymmetry of opinion radicalisation. In particular, the model demonstrates the phenomena of pronounced conservatism versus mild liberalism when individuals are exposed to opposing views, mirroring recent findings on opinion polarisation via social media exposure. Advancing this model, we establish a robust framework that integrates elements from physics, psychology, behavioural science, decision-making theory, and philosophy. We also emphasise the inherent advantages of the quantum approach over traditional models, suggesting a number of new directions for future research work on quantum-mechanical models of human cognition and decision-making.

COVID-19 and Suicide Tendency: Prediction and Risk Factor Analysis Using Machine Learning and Explainable AI

INTRODUCTION: Pandemics and epidemics have frequently led to a significant increase in the suicide rate in affected regions. However, these unnecessary deaths can be prevented by identifying the risk factors and intervening earlier with those at risk. Numerous empirical studies have exhaustively documented multiple suicide risk factors. In addition, many evidence-based approaches have employed machine learning models to diagnose vulnerable groups, a task that would otherwise be challenging if only human cognition were employed. To date, to the best of our knowledge, no research has been conducted on COVID-19-related suicide prediction.OBJECTIVES: This research, aims to develop a machine-learning model capable of identifying individuals who are contemplating suicide due to COVID-19-related complexities and assessing the potential risk factors.METHODS: We trained a gradient-boosting model based on tree-based learners on 10067 data consisting of 76 features, which were primarily responses to socio-demographic, behavioural, and psychological questions about COVID-19 and suicidal behaviours.RESULTS: The final model predicted individuals at risk with an auROC score of 0.77 and a 95% confidence interval of 0.77 to 0.88. The optimal cutoff produced a sensitivity of 31.37 percent and a specificity of 82.35 percent in predicting suicidal tendencies. However, the auPRC was only 0.26, with a 95 percent confidence interval of 0.13 to 0.38, as the class distribution was extremely unbalanced. Consequently, the scores for precision and recall were 0.35 and 0.31, respectively.CONCLUSION: We investigated the risk factors, the majority of which were associated with sleeping difficulties, fear of COVID-19, social interactions, and other socio-demographic factors. The identified risk factors can be considered when formulating a policy to prevent COVID-19-related suicides, which can impose a long-term economic and health burden on society.

Type I interferon signaling, cognition and neurodegeneration following COVID-19: update on a mechanistic pathogenetic model with implications for Alzheimer's disease.

COVID-19's effects on the human brain reveal a multifactorial impact on cognition and the potential to inflict lasting neuronal damage. Type I interferon signaling, a pathway that represents our defense against pathogens, is primarily affected by COVID-19. Type I interferon signaling, however, is known to mediate cognitive dysfunction upon its dysregulation following synaptopathy, microgliosis and neuronal damage. In previous studies, we proposed a model of outside-in dysregulation of tonic IFN-I signaling in the brain following a COVID-19. This disruption would be mediated by the crosstalk between central and peripheral immunity, and could potentially establish feed-forward IFN-I dysregulation leading to neuroinflammation and potentially, neurodegeneration. We proposed that for the CNS, the second-order mediators would be intrinsic disease-associated molecular patterns (DAMPs) such as proteopathic seeds, without the requirement of neuroinvasion to sustain inflammation. Selective vulnerability of neurogenesis sites to IFN-I dysregulation would then lead to clinical manifestations such as anosmia and cognitive impairment. Since the inception of our model at the beginning of the pandemic, a growing body of studies has provided further evidence for the effects of SARS-CoV-2 infection on the human CNS and cognition. Several preclinical and clinical studies have displayed IFN-I dysregulation and tauopathy in gene expression and neuropathological data in new cases, correspondingly. Furthermore, neurodegeneration identified with a predilection for the extended olfactory network furthermore supports the neuroanatomical concept of our model, and its independence from fulminant neuroinvasion and encephalitis as a cause of CNS damage. In this perspective, we summarize the data on IFN-I as a plausible mechanism of cognitive impairment in this setting, and its potential contribution to Alzheimer's disease and its interplay with COVID-19.

The non-Cattell-Horn-Carroll (non-CHC) Model of Ancillary Broad and Narrow Abilities

The Cattell-Horn-Carroll (CHC) theory of abilities is a comprehensive model of human cognitive abilities denoted by the G-code. Developed by three prominent psychologists - Raymond Cattell, John Horn, and John Carroll - the CHC theory suggests that there are three strata of cognitive abilities. In this paper, the authors argued that the CHC model faces seven challenges: First, it is a complex model with numerous factors and subtests, posing difficulties in administration and interpretation. Second, the theory itself has limited coverage and may not address all cognitive abilities. Third, critics have been arguing that the CHC theory has failed to sufficiently address socio-cultural variations. Fourth, the CHC model of broad and narrow abilities lacks concrete operational definitions, relying on theoretical constructs. Fifth, it always fails to account for complex interactions between cognitive abilities. Sixth, the CHC model continually evolves, requiring frequent updates. Lastly, the CHC theory might not fully capture individual differences in cognitive abilities and potential, leaving room for improvement. There are alternative models (e.g., extended fluid-crystallized and verbal-perceptual-image rotation theories) to the CHC model. Despite these limitations, the CHC model remains a valuable tool for understanding cognitive abilities. In this paper, the authors have chosen to focus on the non-CHC categories of broad abilities with their respective narrow abilities. Termed as ancillary abilities denoted by the Q-code, this paper aims to provide another model of abilities (not necessarily cognitive), not to replace the current CHC model, but to complement or supplement it.

Structural-functional brain network coupling predicts human cognitive ability

Individual differences in general cognitive ability (GCA) have a biological basis within the structure and function of the human brain. Network neuroscience investigations revealed neural correlates of GCA in structural as well as in functional brain networks. However, whether the relationship between structural and functional networks, the structural-functional brain network coupling (SC-FC coupling), is related to individual differences in GCA remains an open question. We used data from 1030 adults of the Human Connectome Project, derived structural connectivity from diffusion weighted imaging, functional connectivity from resting-state fMRI, and assessed GCA as a latent g-factor from 12 cognitive tasks. Two similarity measures and six communication measures were used to model possible functional interactions arising from structural brain networks. SC-FC coupling was estimated as the degree to which these measures align with the actual functional connectivity, providing insights into different neural communication strategies. At the whole-brain level, higher GCA was associated with higher SC-FC coupling, but only when considering path transitivity as neural communication strategy. Taking region-specific variations in the SC-FC coupling strategy into account and differentiating between positive and negative associations with GCA, allows for prediction of individual cognitive ability scores in a cross-validated prediction framework (correlation between predicted and observed scores: r = 0.25, p < .001). The same model also predicts GCA scores in a completely independent sample (N = 567, r = 0.19, p < .001). Our results propose structural-functional brain network coupling as a neurobiological correlate of GCA and suggest brain region-specific coupling strategies as neural basis of efficient information processing predictive of cognitive ability.

Can Robots Be Smarter than Human Beings?

The rapid advancements in artificial intelligence (AI) and robotics have fuelled speculation about the possibility of machines surpassing human intelligence. This abstract explores the question of whether robots can become smarter than human beings. By analyzing the capabilities of contemporary AI technologies and understanding the fundamental characteristics of human intelligence, we assess the potential for machines to outpace human cognitive abilities. To address this question, we examine the current state of AI development, including machine learning, deep learning, and neural networks. We discuss how these technologies enable robots to process vast amounts of data, learn from it, and make autonomous decisions. Additionally, we explore the concept of "general intelligence" and its significance in comparing human intelligence to AI. While machines excel in narrow domains and specific tasks, achieving true human-like general intelligence remains an ongoing challenge. Furthermore, the abstract delves into the unique aspects of human intelligence that set it apart from current AI capabilities. Human cognition encompasses various complex abilities such as creativity, emotional intelligence, moral reasoning, and common sense understanding, which have proven difficult to replicate in machines. We examine the limitations and challenges faced by AI in these areas and explore the philosophical and ethical implications of creating machines that surpass human cognitive capacities. Ultimately, the abstract concludes that while robots have shown remarkable progress in specific domains, surpassing human intelligence in its entirety is a complex and elusive goal. Human intelligence encompasses a multitude of intricacies that extend beyond pure computational power. However, the abstract acknowledges the potential for future advancements in AI and robotics to challenge the boundaries of human cognitive abilities, warranting continued research and thoughtful consideration of their societal impact.

Spatial–temporal hypergraph based on dual-stage attention network for multi-view data lightweight action recognition

For the problems of irrelevant frames and high model complexity in action recognition, we propose a Spatial–Temporal Hypergraph based on Dual-Stage Attention Network (STHG-DAN) for multi-view data lightweight action recognition. It includes two stages: Temporal Attention Mechanism based on Trainable Threshold (TAM-TT) and Hypergraph Convolution based on Dynamic Spatial–Temporal Attention Mechanism (HG-DSTAM). In the first stage, TAM-TT uses a learning threshold to extract keyframes from multi-view videos, with the multi-view data serving as a guarantee for providing more comprehensive information subsequently; In the second stage, HG-DSTAM divides the human joints into three parts: trunk, hand and leg to build spatial–temporal hypergraphs, extracts high-order features from spatial–temporal hypergraphs constructed of multi-view human body joints, inputs them into the dynamic spatial–temporal attention mechanism, and learns the intra frame correlation of multi-view data between the joint features of body parts, which can obtain the significant areas of action; We use multi-scale convolution operation and depth separable network, which can realize efficient action recognition with a few trainable parameters. We experiment on the NTU-RGB+D, NTU-RGB+D 120 and the imitating traffic police gesture dataset. The performance and accuracy of the model are better than the existing algorithms, effectively improving the machine and human body language interaction cognitive ability.

Hydrocortisone reduces altruistic punishment in healthy men

Psychosocial stress modulates social cognition and behavior in humans. One potentially mediating factor is cortisol as part of the human endocrine stress response. With a double-blind, placebo-controlled between-subject study design, we tested possible dose-dependent effects of hydrocortisone (0 mg, 5 mg and 20 mg) in 85 healthy males. During a socio-economic decision-making task we measured trust, trustworthiness, sharing, punishment, and non-social risk behavior. Social value orientation (SVO) was also assessed. We observed significantly lower levels of punishment after hydrocortisone, especially in the 20 mg group. Drug-induced salivary cortisol correlated negatively with punishment behavior. None of the other facets of social behavior or the SVO were affected by hydrocortisone. Our results suggest that hydrocortisone reduces the propensity to punish unfair behavior. Future studies are needed to further disentangle the role played by various psychobiological mechanisms within the stress response as well as their complex interplay on social behavior and cognition.

Differential effects of bilateral hippocampal CA3 damage on the implicit learning and recognition of complex event sequences

ABSTRACT Learning regularities in the environment is a fundament of human cognition, which is supported by a network of brain regions that include the hippocampus. In two experiments, we assessed the effects of selective bilateral damage to human hippocampal subregion CA3, which was associated with autobiographical episodic amnesia extending ~50 years prior to the damage, on the ability to recognize complex, deterministic event sequences presented either in a spatial or a non-spatial configuration. In contrast to findings from related paradigms, modalities, and homologue species, hippocampal damage did not preclude recognition memory for an event sequence studied and tested at four spatial locations, whereas recognition memory for an event sequence presented at a single location was at chance. In two additional experiments, recognition memory for novel single-items was intact, whereas the ability to recognize novel single-items in a different location from that presented at study was at chance. The results are at variance with a general role of the hippocampus in the learning and recognition of complex event sequences based on non-adjacent spatial and temporal dependencies. We discuss the impact of the results on established theoretical accounts of the hippocampal contributions to implicit sequence learning and episodic memory.

Dialogical cognition

In this article we review Per Linell's work within the last five decades that led to his dialogism framework, which he defines as a general epistemology of language, cognition and communication. We critically discuss how his contribution on the one hand, altered and qualified existent models within language, communication and cognitive science, because dialogism removed language and cognition from their abstract and mental seat in the brain, and embedded them instead in situational contexts and embodied interaction. In that sense, his dialogism successfully replaced monological assumptions about the mind, action and thinking with more contextual and temporally distributed ones. On the other hand, we also question why Linell has not pursued a more rigorous empirical program for studying human cognition, when he did establish a theoretical apparatus for approaching cognition from a dialogical starting point. In going through Linell's arguments over the past five decades we suggest that this absence of an empirical program is due to his humanistic roots which both have sensitised him to appreciating the contingencies and dynamics of human sense making and cognition, and have impeded him from buying into a necessary condition for pursuing a cognitive analysis, even if he conceptually and methodologically accepts a distributed view on cognition. The outcome of this discussion leads to an empirical-based cognitive analysis of a medical interaction. Altogether, the purpose of this article is to show how Linell's conceptual framework can be put to use in ways that make a dialogical cognitive science achievable.

A MAGDM approach for evaluating the impact of artificial intelligence on education using 2-tuple linguistic q-rung orthopair fuzzy sets and Schweizer-Sklar weighted power average operator.

The impact of artificial intelligence (AI) in education can be viewed as a multi-attribute group decision-making (MAGDM) problem, in which several stakeholders evaluate the advantages and disadvantages of AI applications in educational settings according to distinct preferences and criteria. A MAGDM framework can assist in providing transparent and logical recommendations for implementing AI in education by methodically analyzing the trade-offs and conflicts among many components, including ethical, social, pedagogical, and technical concerns. A novel development in fuzzy set theory is the 2-tuple linguistic q-rung orthopair fuzzy set (2TLq-ROFS), which is not only a generalized form but also can integrate decision-makers quantitative evaluation ideas and qualitative evaluation information. The 2TLq-ROF Schweizer-Sklar weighted power average operator (2TLq-ROFSSWPA) and the 2TLq-ROF Schweizer-Sklar weighted power geometric (2TLq-ROFSSWPG) operator are two of the aggregation operators we create in this article. We also investigate some of the unique instances and features of the proposed operators. Next, a new Entropy model is built based on 2TLq-ROFS, which may exploit the preferences of the decision-makers to obtain the ideal objective weights for attributes. Next, we extend the VIseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR) technique to the 2TLq-ROF version, which provides decision-makers with a greater space to represent their decisions, while also accounting for the uncertainty inherent in human cognition. Finally, a case study of how artificial intelligence has impacted education is given to show the applicability and value of the established methodology. A comparative study is carried out to examine the benefits and improvements of the developed approach.

The beneficial role of curiosity on route memory in children

IntroductionThere has been a growing interest in the role of innate curiosity on facets of human cognition, such as in spatial learning and memory. Yet, it is unclear how state level curiosity evoked by the current environment could interact differentially with trait curiosity, to impact spatial memory performance.MethodsWe assessed the influence of trait and state curiosity on route memory. Forty-two 10-year-old children with low and high-trait curiosity (20 Females; 22 Males) actively explored virtual environments that elicited varying levels of uncertainty (i.e., state-curiosity).ResultsAs trait curiosity increased, so did memory performance in low and high uncertainty conditions, suggesting that high-curiosity children can better recruit cognitive resources within non-optimal environments. Children with high compared to low trait curiosity also reported greater feelings of presence during exploration. Importantly, in environments with medium uncertainty, children with low trait curiosity were able to perform as well as those with high curiosity.DiscussionResults show that individual differences in trait curiosity influence route learning and these interact dynamically with state-curiosity invoked within different environments.

Exploring factors in Human Cognition that affect Human Performance during Takeover in Autonomous Driving

Driver's mental load is an important influence on the design of the human-machine interface (HMI) for self-driving cars. During the autopilot-forced takeover process, aspects of the user's cognition, decision-making, attention, and emotional state affect the takeover performance. Therefore, the cognitive and psychological states of drivers must be considered in the HMI design of self-driving cars. And a good HMI design will reduce the user's mental burden and improve the takeover performance. In this study, a model affecting drivers' takeover performance was designed through human factors. And we refine the analysis from three aspects of cognitive load, emotional load, and environmental load by making subjective likert-style-questionnaires, derive the main influencing factors through the reliability analysis of the data, factor analysis, and other methods. This work finds that the clarity, comprehensiveness, and level of humanization of takeover messages, as well as the driver's arousal level when receiving the message, have the greatest impact on the driver's workload.