Eye Gaze Behavior Research Articles

Enhancing social robot's direct gaze expression through vestibulo-ocular movements

Non-verbal communication, especially eye contact, plays a crucial role in human interaction. Integrating eye contact capabilities into robotic behavior presents challenges that require comprehensive interdisciplinary collaboration from the fields of vision science and robotics. This study introduces an innovative methodology to improve the user's perception of the tabletop social robot Haru's direct gaze by explaining how eye and head movements can be managed through simulating visual perception processing. By considering the mechanisms of eye and head movements during visual perception processing, the robot's eye and head gaze behaviors are intended to replicate human gaze movements with Vestibulo-Ocular Movement (VOM) during face-to-face interaction, ultimately enhancing the user's perception of the robot's direct gaze.

The Gaze Dialogue Model: Nonverbal Communication in HHI and HRI.

When humans interact with each other, eye gaze movements have to support motor control as well as communication. On the one hand, we need to fixate the task goal to retrieve visual information required for safe and precise action-execution. On the other hand, gaze movements fulfil the purpose of communication, both for reading the intention of our interaction partners, as well as to signal our action intentions to others. We study this Gaze Dialogue between two participants working on a collaborative task involving two types of actions: 1) individual action and 2) action-in-interaction. We recorded the eye-gaze data of both participants during the interaction sessions in order to build a computational model, the Gaze Dialogue, encoding the interplay of the eye movements during the dyadic interaction. The model also captures the correlation between the different gaze fixation points and the nature of the action. This knowledge is used to infer the type of action performed by an individual. We validated the model against the recorded eye-gaze behavior of one subject, taking the eye-gaze behavior of the other subject as the input. Finally, we used the model to design a humanoid robot controller that provides interpersonal gaze coordination in human-robot interaction scenarios. During the interaction, the robot is able to: 1) adequately infer the human action from gaze cues; 2) adjust its gaze fixation according to the human eye-gaze behavior; and 3) signal nonverbal cues that correlate with the robot's own action intentions.

Effects of In-Vehicle Touchscreen Location on Driver Task Performance, Eye Gaze Behavior, and Workload During Conditionally Automated Driving: Nondriving-Related Task and Take-Over.

This study investigated the effects of nondriving-related task (NDRT) touchscreen location and NDRT difficulty level on the driver task performance, eye gaze behavior, and workload during SAE Level 3 conditionally automated driving. Two driver tasks were considered: a visuomanual NDRT and a take-over task. Touchscreens are expected to play important roles inside automated vehicles. However, few studies have investigated the driver-touchscreen interaction during automated driving. A driving simulator experiment was conducted. The experimental task consisted of two successive subtasks: an NDRT followed by a take-over task. NDRT touchscreen location (Upper Left, Upper Right, and Lower Right) and NDRT difficulty level (Easy and Hard) were the independent variables. A set of driver task performance, eye gaze behavior, and perceived workload measures were employed for each subtask as the dependent variables. NDRT touchscreen location significantly affected both the NDRT and the take-over task performance. Lower Right was superior to Upper Right in the NDRT performance but was inferior in the take-over task performance. NDRT touchscreen location affected the perceived physical workload of the NDRT. NDRT difficulty level affected the perceived workload of the take-over task. The research findings enhance our understanding of how NDRT touchscreen location and NDRT difficulty level impact the driver task performance during conditionally automated driving, and, further provide useful design implications and knowledge. The study results would inform the NDRT touchscreen interface design and the NDRT design for conditionally automated vehicles.

Basketball free-throw training with augmented reality-based optimal shot trajectory for novice shooters

We propose an augmented reality (AR)-based training system for basketball free-throws. The optimal shot trajectory for free-throws is projected by a head-mounted display according to the shooter’s release point. The efficacy of the training system was assessed in novice shooters by comparing changes in success rates and eye-gaze behavior (quiet eye [QE]) between AR-training and control-training groups. The success rate during the AR training with the optimal trajectory did not differ from the pre-training rate; however, in post-AR training, i.e., after removal of the optimal trajectory, the success rate increased. Additionally, AR training increased the QE duration (QED) compared with that recorded during pre- and post-training blocks. In contrast, the control group showed no change in the success rate or QED. These findings imply that our AR training system affected QE behavior and improved free-throwing shooting performance after training. Thus, our system is expected to enhance basketball free-throw shooting performance.

Variations in Eye-Gaze Behaviors Evident When Five Adults With Aphasia-Based Alexia Read Multisentence Passages.

The problems that people with aphasia encounter when reading passages are poorly understood. This study's purpose was in-depth examination of eye-gaze behaviors exhibited by five people with aphasia-based alexia. Five adults with aphasia-based alexia and five neurotypical adults (NAs) read paragraphs while having their eye movements recorded. Acquired data included descriptive characterization of overall eye-gaze behaviors and determination of the (a) percent of fixated words, (b) average fixation duration, (c) average initial and total summed fixation durations of processing attempts on individual words, and (d) effects of word length and frequency on fixation durations. Careful examination of these data allowed examination of consistencies and discrepancies among people with aphasia and supported speculation about underlying deficits. Case participants exhibited unique fixation behaviors in comparison to one another and to neurotypical adults. Case participants' total reading time, percent of fixated words, average fixation duration, and average initial and total summed fixation durations on passage words exceeded those of neurotypical adults. Four of five exhibited positive word length and negative word frequency correlations with fixation durations. People with aphasia display eye-gaze behaviors unique to them and differing from those of NAs when reading texts. Better understanding of the connection between specific eye-gaze behaviors and problems decoding words, linking them with lexical-semantic information, and constructing meaning from written content is necessary to further progress in developing effective assessment tools and treatments.

The impact of eye gaze on social interactions of females in virtual reality: The mediating role of the uncanniness of avatars and the moderating role of task type

The introduction of the Metaverse and the rise of social virtual reality platforms such as VRChat has led to increased communication and collaboration in virtual reality. As interactions in social virtual reality take place through avatars, the behaviour, and in particular the eye gaze of an avatar can have an impact on the user experience. However, it is still unclear which features of the user experience are most influenced. Therefore, this study used data from 44 female participants to investigate how avatar gaze behaviour (static eyes vs. real-time tracked eyes) affects their perceived quality of communication and the amount of one-sided eye contact participants make during a communicative or collaborative task in virtual reality. In addition, the study investigated the mediating role of perceived avatar “uncanniness” (i.e., the finding that humanoid objects that imperfectly resemble real people evoke feelings of discomfort) in this relationship and the moderation effect of two multi-user scenarios (collaboration vs. communication). The results showed that uncanniness directly affected the perceived quality of communication. However, it did not significantly mediate the relationship between avatar eye gaze behaviour and the quality of communication or the amount of one-sided eye contact. Finally, there were no significant differences in user experience between the two scenarios. From this, we can conclude that the uncanniness of an avatar being interacted with in VR is not enough to hinder communication and collaboration in an immersive medium. And even if an avatar is perceived as uncanny, normative communication cues such as eye contact are still present. Notably, due to sample availability, the results of this study are based on a female-only sample. Thus, future research can benefit from exploring the outlined effects in a more gender-balanced sample.

Evaluation of Eye Gaze Dynamics During Physician-Patient-Computer Interaction in Federally Qualified Health Centers: Systematic Analysis.

Understanding the communication between physicians and patients can identify areas where they can improve and build stronger relationships. This led to better patient outcomes including increased engagement, enhanced adherence to treatment plan, and a boost in trust. This study investigates eye gaze directions of physicians, patients, and computers in naturalistic medical encounters at Federally Qualified Health Centers to understand communication patterns given different patients' diverse backgrounds. The aim is to support the building and designing of health information technologies, which will facilitate the improvement of patient outcomes. Data were obtained from 77 videotaped medical encounters in 2014 from 3 Federally Qualified Health Centers in Chicago, Illinois, that included 11 physicians and 77 patients. Self-reported surveys were collected from physicians and patients. A systematic analysis approach was used to thoroughly examine and analyze the data. The dynamics of eye gazes during interactions between physicians, patients, and computers were evaluated using the lag sequential analysis method. The objective of the study was to identify significant behavior patterns from the 6 predefined patterns initiated by both physicians and patients. The association between eye gaze patterns was examined using the Pearson chi-square test and the Yule Q test. The results of the lag sequential method showed that 3 out of 6 doctor-initiated gaze patterns were followed by patient-response gaze patterns. Moreover, 4 out of 6 patient-initiated patterns were significantly followed by doctor-response gaze patterns. Unlike the findings in previous studies, doctor-initiated eye gaze behavior patterns were not leading patients' eye gaze. Moreover, patient-initiated eye gaze behavior patterns were significant in certain circumstances, particularly when interacting with physicians. This study examined several physician-patient-computer interaction patterns in naturalistic settings using lag sequential analysis. The data indicated a significant influence of the patients' gazes on physicians. The findings revealed that physicians demonstrated a higher tendency to engage with patients by reciprocating the patient's eye gaze when the patient looked at them. However, the reverse pattern was not observed, suggesting a lack of reciprocal gaze from patients toward physicians and a tendency to not direct their gaze toward a specific object. Furthermore, patients exhibited a preference for the computer when physicians directed their eye gaze toward it.

Exploring the relationship of working memory to the temporal distribution of pausing and revision behaviors during L2 writing

AbstractThis study examined the extent to which L2 writers with varied working memory display differential pausing and revision behaviors at different periods during writing. The participants were 30 advanced Chinese L2 users of English, who wrote an argumentative essay. While composing, participants’ keystrokes and eye-gaze movements were recorded to capture their pausing, revision, and eye-gaze behaviors. The working memory battery included tests of phonological and visual short-term memory and executive functions. We divided the writing process into five equal periods. The results revealed that participants’ pausing and revision patterns were consistent with previous findings that planning, linguistic encoding, and monitoring processes dominate the initial, middle, and later composing periods, respectively. Various working memory components had differential effects on pausing depending on period, largely reflecting the predictions of Kellogg’s (1996, 2001) model. However, we identified no differences in the temporal distribution of revision behaviors contingent on working memory.

Examining speech-based phonological recoding during reading for adolescent deaf signers.

Much of the debate regarding literacy development in deaf and hard-of-hearing readers surrounds whether there is dependence on phonological decoding of print to speech for such readers, and the literature is mixed. While some reports of deaf children and adults demonstrate the influence of speech-based processing during reading, others find little to no evidence of speech-sound activation. In order to examine the role of speech-based phonological codes when reading, we utilized eye-tracking to examine eye-gaze behaviors employed by deaf children and a control group of hearing primary-school children when encountering target words in sentences. The target words were of three types: correct, homophonic errors, and nonhomophonic errors. We examined eye-gaze fixations when first encountering target words and, if applicable, when rereading those words. The results revealed that deaf and hearing readers differed in their eye-movement behaviors when re-reading the words, but they did not demonstrate differences for first encounters with the words. Hearing readers treated homophonic and nonhomophonic error words differently during their second encounter with the target while deaf readers did not, suggesting that deaf signers did not engage in phonological decoding to the same degree as hearing readers did. Further, deaf signers performed fewer overall regressions to target words than hearing readers, suggesting that they depended less on regressions to resolve errors in the text. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

Driver behaviors assisted by different human machine interfaces to avoid rear-end collisions during level 2 automated driving

Objective To practically apply level 2 automated driving in complex traffic conditions, it is necessary to prompt driver behaviors to prevent potential accidents in areas where manual interventions are frequently required. Methods A driving simulator experiment with 20 participants was conducted to evaluate the impact of different human machine interfaces (HMIs) on drivers’ interventions in terms of braking to avoid rear-end collisions during level 2 automated driving when a motorcycle abruptly cut in near intersections. Two types of HMIs were tested: a static HMI that informed drivers about approaching intersections, and a sensor HMI that displayed real-time object recognition results. Each driver participated in five experimental conditions, which varied the presence or absence of the static and sensor HMIs during level 2 automated driving, with manual driving serving as the baseline condition. Results The maximum deceleration in terms of braking to avoid rear-end collisions was significantly larger when level 2 automated driving was used without any HMI, compared to that of manual driving. However, when the sensor HMI was applied together with the static HMI during level 2 automated driving, a comparable time to collision could be achieved with a significantly smaller deceleration, compared to that without any HMI. Drivers’ eye-gaze behaviors revealed that no significant difference existed in the percentages of gaze to the road center area, indicating that they were not distracted by the HMIs. Finally, drivers’ attention levels to surrounding traffic and feeling of safety were significantly higher when level 2 automated driving was used in combination with the static and sensor HMIs. Conclusions The results demonstrated that the combination of static and sensor HMIs successfully aided drivers in ensuring driving safety with a significantly smaller deceleration to avoid rear-end collisions during level 2 automated driving. Furthermore, drivers’ attention levels were maintained, and their feeling of safety was improved when both HMIs were used in combination.

A Platform for Tracking Surgeon and Observer Gaze as a Surrogate for Attention in Ophthalmic Surgery.

To develop and validate a platform that can extract eye gaze metrics from surgeons observing cataract and vitreoretinal procedures and to enable post hoc data analysis to assess potential discrepancies in eye movement behavior according to surgeon experience. Experimental, prospective, single-center study. Eleven ophthalmic surgeons observing deidentified vitreoretinal and cataract surgical procedures performed at a single university-based medical center. An open-source platform was developed to extract gaze coordinates and metrics from ophthalmic surgeons via a computer vision algorithm in conjunction with a neural network to track and segment instruments and tissues, identifying areas of attention in the visual field of study subjects. Eleven surgeons provided validation data by watching videos of 6 heterogeneous vitreoretinal and cataract surgical phases. Accuracy and distance traveled by the eye gaze of participants and overlap of the participants' eye gaze with instruments and tissues while observing surgical procedures. The platform demonstrated repeatability of > 94% when acquiring the eye gaze behavior of subjects. Attending ophthalmic surgeons and clinical fellows exhibited a lower overall cartesian distance traveled in comparison to resident physicians in ophthalmology (P<0.02). Ophthalmology residents and clinical fellows exhibited more fixations to the display area where surgical device parameters were superimposed than attending surgeons (P < 0.05). There was a trend toward gaze overlap with the instrument tooltip location among resident physicians in comparison to attending surgeons and fellows (41.42% vs. 34.8%, P>0.2). The number and duration of fixations did not vary substantially among groups (P>0.3). The platform proved effective in extracting gaze metrics of ophthalmic surgeons. These preliminary data suggest that surgeon gaze behavior differs according to experience.

Driver eye glance behavior and performance with camera-based visibility systems versus mirrors

Objective Drivers’ ability to extract visual information efficiently from mirrors or camera-based visibility systems impacts driving performance when carrying out maneuvers such as lane changes. The objective of the research was to compare drivers’ eye gaze behavior and driving performance with mirrors versus camera-based visibility systems (i.e., CMS, or camera monitor system) to identify any differences and possible impacts on safety. Methods A test track study was conducted comparing drivers’ eye gaze and lane change behavior when driving a vehicle equipped with outside mirrors versus a prototype CMS. Participants’ opinions regarding usability and comfort in using mirrors versus the tested CMS were also obtained using a post-drive questionnaire. Results Study results were somewhat mixed but did demonstrate that with the tested CMS, participants took longer to pass a slower moving vehicle and maintained a greater resultant distance from the passed vehicle. Additionally, participants had a greater number of fixations to the CMS displays compared to the outside rearview mirrors. Results also found slight perceived advantages for the tested CMS in regard to ease of use, comfortability, and visibility. When asked to choose which rear visibility technology they would prefer to use in everyday driving, most participants preferred the outside rearview mirrors over the tested prototype CMS or having both systems. However, not all lane change and gaze metrics followed the same pattern. Conclusions In this study, participants’ longer time to pass a slower moving vehicle, greater distance when passing a slower moving vehicle, greater number of fixations, and lower subjective ratings with the tested CMS may indicate difficulty in judging distances and focusing on the electronic image. This study provides preliminary findings that suggest differences in driving behavior exist between a single tested prototype CMS and outside rearview mirrors and is a foundational step toward evaluating whether these trends are consistent across different systems and overall implications for safe driving behavior.

Impact of duration of monitoring before takeover request on takeover time with insights into eye tracking data

Safety has become the primary concern of automated driving system (ADS) in recent years. Compared with highly automated driving (L4 and above), conditionally automated driving (L3/L3+ ADS) seems to be a moderate choice, where drivers are required to respond to the takeover request (TOR) whenever necessary. It is the system’s responsibility to make sure that the takeovers would be safe at the time of issuing the TOR. To realize that, a lot of factors need to be taken into consideration. As it has been found that drivers’ eyes-on-road gazes increase slowly in the first few seconds while transferring to manual driving from automated driving and drivers’ gaze behaviors are related with situation awareness, the main aim of this study is to investigate the impact of duration of monitoring before the TOR on takeover time and whether there is a positive or negative relationship between the two. To verify these, we designed a driving simulator study where the TOR was issued 0 s, 5 s, 10 s and ≥ 15 s after the non-driving-related task has ended. Twelve scenarios were designed, and the results from 36 participants showed that there was indeed a statistically significant difference, however, the relationship was neither positive nor negative, which was close to a parabola. Analyzing results of eye movements and gaze behavior further supported this conclusion. It is therefore concluded the duration of monitoring before the TOR should neither be too short nor too long, and 5–7 s would be appropriate choices. This is desirable not only for improving takeover performance of drivers but also for improving the prediction model for predicting takeover performance of drivers that has yet to be studied, so as to improve safety, reliability and acceptance of the ADS.

ACT-R based human digital twin to enhance operators' performance in process industries.

To ensure safe and efficient operation, operators in process industries have to make timely decisions based on time-varying information. A holistic assessment of operators' performance is, therefore, challenging. Current approaches to operator performance assessment are subjective and ignore operators' cognitive behavior. In addition, these cannot be used to predict operators' expected responses during novel situations that may arise during plant operations. The present study seeks to develop a human digital twin (HDT) that can simulate a control room operator's behavior, even during various abnormal situations. The HDT has been developed using the ACT-R (Adaptive Control of Thought-Rational) cognitive architecture. It mimics a human operator as they monitor the process and intervene during abnormal situations. We conducted 426 trials to test the HDT's ability to handle disturbance rejection tasks. In these simulations, we varied the reward and penalty parameters to provide feedback to the HDT. We validated the HDT using the eye gaze behavior of 10 human subjects who completed 110 similar disturbance rejection tasks as that of the HDT. The results indicate that the HDT exhibits similar gaze behaviors as the human subjects, even when dealing with abnormal situations. These indicate that the HDT's cognitive capabilities are comparable to those of human operators. As possible applications, the proposed HDT can be used to generate a large database of human behavior during abnormalities which can then be used to spot and rectify flaws in novice operator's mental models. Additionally, the HDT can also enhance operators' decision-making during real-time operation.

An Eye Tracking Investigation of Young People’s Gaze Behaviour to Gambling and Non-Gambling Moving Adverts

Introduction: Data from several areas of public health (e.g., harmful alcohol and tobacco) are consistent with the assertion that children’s exposure to advertising strategies increases intention to consume such products. Most studies have measured self-rated impact of gambling advertising using questionnaires. Given that gambling advertisements come in different forms such as print media/television advertising and contain variable content, it is difficult to understand using subjective measures which aspects of the gambling advertisements increase craving and desire to trigger a gambling session. In the present study, we applied a novel data-driven methodology that directly tracks eye movements to reveal attentional biases towards gambling adverts and promotions by examining differences in young people’s eye gaze behaviour when watching gambling and non-gambling (control) moving adverts. Method: A total of 98 (16–18 years old) children who self-identify as having a low or high craving to gambling watched gambling and non-gambling (control) television adverts, while their eye movements were recorded. Results: The results show that the data-driven method can isolate video clips that best distinguish people on the low-high craving spectrum, reveal the type of each video clip with the largest group differences, and accurately predict young people’s gambling craving on the basis of eye movement patterns. Conclusion: Our findings demonstrate that young people’s craving for gambling can be predicted based on their eye movements to video clips of gambling advertisements and that certain features of gambling advertisements may be more appealing to some group of viewers, particularly those with high craving for gambling.

Detection of Reading Impairment from Eye-Gaze Behaviour using Reinforcement Learning

Experimental psychology and neuroscience reveal that decision-behavior plays a dominant role in human-selective-attention when it comes to reading, object and scene detection. Difficulties in reading are easily reflected by eye-movement patterns. Hence, modelling eye-gaze behaviour for normal readers and people with reading impairments can greatly help in contrasting reading strategies used, which can in turn help in early identification and diagnosis of impairments such as dyslexia. This paper introduces a novel method of formulating a reinforcement learning model that is explainable, and can obtain the sequence of gaze targets based on recorded observations of dyslexic and non-dyslexic children. Results reveal that despite being a less sophisticated model, it is able to obtain the optimal reading policy of the ideal reader, from a set of good and poor readers with the help of a strong reward system and Q-Learning agent.

Looking deep in the eyes: Investigating interpretation methods for neural models on reading tasks using human eye-movement behaviour

This paper provides the first broad overview of the relation between different interpretation methods and human eye-movement behaviour across different tasks and architectures. The interpretation methods of neural networks provide the information the machine considers important, while the human eye-gaze has been believed to be a proxy of the human cognitive process. Thus, comparing them explains machine behaviour in terms of human behaviour, leading to improvement in machine performance through minimising their difference. We consider three types of natural language processing (NLP) tasks: sentiment analysis, relation classification and question answering, and four interpretation methods based on: simple gradient, integrated gradient, input-perturbation and attention, and three architectures: LSTM, CNN and Transformer. We leverage two corpora annotated with eye-gaze information: the Zuco dataset and the MQA-RC dataset. This research sets up two research questions. First, we investigate whether the saliency (importance) of input-words conform with those from human eye-gaze features. To this end, we compute a saliency distance (SD) between input words (by an interpretation method) and an eye-gaze feature. SD is defined as the KL-divergence between the saliency distribution over input words and an eye-gaze feature. We found that the SD scores vary depending on the combinations of tasks, interpretation methods and architectures. Second, we investigate whether the models with good saliency conformity to human eye-gaze behaviour have better prediction performances. To this end, we propose a novel evaluation device called “SD-performance curve” (SDPC) which represents the cumulative model performance against the SD scores. SDPC enables us to analyse the underlying phenomena that were overlooked using only the macroscopic metrics, such as average SD scores and rank correlations, that are typically used in the past studies. We observe that the impact of good saliency conformity between humans and machines on task performance varies among the combinations of tasks, interpretation methods and architectures. Our findings should be considered when introducing eye-gaze information for model training to improve the model performance.

Eye-Tracking Active Indicators of Insider Threats: Detecting Illicit Activity During Normal Workflow

Cyber challenges faced by organizations today involve malicious inside actors, often labeled insider threats (ITs). These present a difficult challenge in that the most well-designed cybersecurity apparatus is vulnerable to those inside the organization who have privileged access to information systems. Innovative methods must be developed to help security analysts narrow the large pool of potential ITs in large organizations to a more manageable number. The purpose of this article is to develop and validate eye-tracking metrics that are diagnostic of IT behavior. Key stimuli, or called active indicator probes, were embedded into a simulated workflow to elicit diagnostic eye-tracking responses. Two environments were simulated: financial and intelligence analysis. We evaluated participants performing as regular workers relative to ITs to identify metrics that distinguished between the two groups. Detection of illicit eye gaze behavior while using chat programs was possible when conversations with accomplices occurred in a separate chat window from normal permissible chat conversations. Validation of results in real work environments is necessary for practical application. However, if the approach proves to translate successfully, automated monitoring of eye-tracking responses may augment existing insider detection methods, within frameworks for best practices in organizational security and cyber defense.

The relationship of proficiency to speed fluency, pausing, and eye-gaze behaviours in L2 writing

In this study we investigated the extent to which writing proficiency predicts L2 writers’ speed fluency, pausing, and eye-gaze behaviours and the cognitive processes that underlie pausing. Additionally, we explored whether these relationships were influenced by stage of writing (beginning, middle stages, end). The participants were 60 Chinese second language users of English, with proficiency levels ranging from CEFR B1 to C1 levels. They all completed two independent TOEFL iBT writing tasks over two sessions, with the tasks being counterbalanced across participants. While composing, we recorded participants’ keystrokes and eye-gaze behaviours. Participants also took part in a stimulated recall session based on the last writing task they had completed. A series of mixed-effects regression models found that proficiency was the strongest predictor of speed fluency. The stimulated recall analysis revealed considerably less variation in cognitive activities among lower- than higher-proficiency writers across writing stages.

Is It Me or the Robot? A Critical Evaluation of Human Affective State Recognition in a Cognitive Task.

A key goal in human-robot interaction (HRI) is to design scenarios between humanoid robots and humans such that the interaction is perceived as collaborative and natural, yet safe and comfortable for the human. Human skills like verbal and non-verbal communication are essential elements as humans tend to attribute social behaviors to robots. However, aspects like the uncanny valley and different technical affinity levels can impede the success of HRI scenarios, which has consequences on the establishment of long-term interaction qualities like trust and rapport. In the present study, we investigate the impact of a humanoid robot on human emotional responses during the performance of a cognitively demanding task. We set up three different conditions for the robot with increasing levels of social cue expressions in a between-group study design. For the analysis of emotions, we consider the eye gaze behavior, arousal-valence for affective states, and the detection of action units. Our analysis reveals that the participants display a high tendency toward positive emotions in presence of a robot with clear social skills compared to other conditions, where we show how emotions occur only at task onset. Our study also shows how different expression levels influence the analysis of the robots' role in HRI. Finally, we critically discuss the current trend of automatized emotion or affective state recognition in HRI and demonstrate issues that have direct consequences on the interpretation and, therefore, claims about human emotions in HRI studies.