Search

We plan to test physics novices and assess their understanding of motion diagrams under instructed valid versus uninstructed valid versus no visual cueing.

A person’s concerns about climate change do not necessarily translate into corresponding actions. In this context, it has been suggested that instilling a sense of urgency leads to more corresponding pro-environmental action (e.g., Brügger et al., 2015; Fesenfeld & Rinscheid, 2021). Moreover it is not only perceived urgency that appears to affect climate change-related beliefs, thoughts, or actions. Climate change is typically framed in and perceived in a temporal context (e.g., Ho, 2021; Kim & Ahn, 2019; Ma et al., 2022; McMichael & Katonivualiku, 2020; Pahl et al., 2014). That is, perceptions and actions related to climate change have been connected to temporal perception. Here, we build on a first experiment which demonstrated severe underestimations of the time (number of years) between two pictures of the same location in the Alps (Pfeuffer & Heckmann, 2023, https://osf.io/jrgk4). We aim at investigating the relationship between the way information on climate change is framed and temporal estimations of changes in the mountain landscape of the Alps in the German population. Furthermore, we investigate how individual climate change anxiety and pro-/anti-environmental attitudes relate to these time estimates. Participants are first confronted with one of five framing texts which either describe climate change objectively, as a disease, as war, or as a race against time or describe a control condition (mountain climate). Then, participants are presented with pairs of pictures that show the same location in the Alps (recent vs. historic). The number of years between the two pictures of a pair varies and participants are asked to estimate how many years have passed between the recent and historic picture. Furthermore, we assess participants' pro-/anti-environmental attitudes via the New Ecological Paradigm Scale (NEP; Dunlap et al., 2000; Schleyer-Lindenmann et al., 2018) and their climate change related anxiety via the Climate Anxiety Scale (CAS; Clayton & Karazsia, 2010; Wullenkord et al., 2021).

Artificial intelligence (AI) solutions are increasingly being used by organizations to select candidates when hiring. The recruitment of highly qualified personnel is an essential task for companies. Selecting the right personnel provides a competitive advantage that determines whether a company has the necessary knowledge, skills, and abilities to succeed in the highly competitive business environment. Especially in times of demographic change and the “War for Talent” (Chambers et al., 1998), companies see the potential of utilizing AI (Acikgoz et al., 2020). On the one hand, this has economic reasons, as AI allows for the fast processing of a large number of applications, which makes applicant selection processes faster and more efficient (Alsever, 2017). On the other hand, AIs ideally contribute to more objectivity and consistency as they process each application in the same way, potentially reducing human-bias in terms of unconscious biases and stereotypes that influence human behavior and decisions (Lacroux & Martin-Lacroux, 2022). At the same time, the use of algorithms also carries the risk of potential discrimination on a large scale, as they can also be vulnerable to biases in terms of gender and ethnicity, for example, if they rely on inaccurate or biased input and training data (Kim, 2017). This can be detrimental to applicants as well as companies. However, in light of the advancing digitization and the shortage of skilled labor, the adoption of AI in recruiting will continue to gain relevance in the coming years (Ahmed, 2018). Employer reputation and branding play an important role in attracting applicants, as the impression gained during the recruitment and selection process influences whether those who initially applied will stay in the applicant pool and ultimately accept a job offer (Acikgoz, 2019). Therefore, considering the applicant's perspective and ensuring that they have a positive opinion of the recruitment process is important for the success of recruitment efforts. Not taking into consideration how potential applicants perceive the use of AI in the applicant selection process can negatively impact the employer’s image, as applicants might, for instance, perceive the use of AI to evaluate humans as demeaning and dehumanizing (Lee, 2018). This attitude of general distrust towards algorithms refers to the phenomenon of algorithm aversion. Research has shown that trust in algorithms remains low, even though many algorithms are superior to humans in certain tasks, because they are more reliable and accurate in most cases (Burton et al., 2020; Jussupow et al., 2020). In addition, several studies have shown that individuals give more weight to human as compared to AI input (Önkal et al., 2009; Promberger & Baron, 2006). A study by Lee (2018) showed that many lay people believe that algorithms are unable to identify good candidates, because they lack the human judgment and intuition that are necessary, in their eyes, for subjective decisions such as hiring employees. This suggests that, overall, trust in AI systems is lower when it comes to assessing people, as it is often the case in Human Resources (HR) (Lacroux & Martin-Lacroux, 2022). The general distrust in algorithms can inhibit the adoption and efficacy of technologies, as trust is a crucial factor for technology acceptance, making it important to establish the right level of trust (Lee, 2018; Lee & See, 2004). Although the phenomenon of algorithm aversion is well established, there are relatively few studies that examine which variables contribute to increasing applicants' trust in algorithmic selection systems and thus decreasing algorithm aversion. Therefore, further research is required for the use of AI and the reduction of algorithm aversion in the HR field. This study examines two factors in terms of their potential influence on trust in AI systems: Personal information inquiries and feedback opportunities. In a paper by Lee (2018), it was suggested for future research to investigate the impact of interaction on the perception of AI decisions. A similar implication for further research was given in an empirical study by Dahm and Dregger (2019), who suggested communicating the score determined by an AI to applicants to increase trust in the AI system. The type of interaction we therefore consider as especially relevant is giving feedback. In this regard, feedback on the applicant’s suitability for a job in the selection process conducted by an AI might impact the applicant’s trust in the AI-based selection system. Yarger et al. (2020) stated in their paper that a growing number of companies are using AI to remove personally identifying information such as name, gender, age, ethnic and racial background as known sources of human and unconscious biases from the hiring process to provide equal opportunity. Blind hiring is a practice increasingly used by organizations to consider applicants primarily on the basis of their skills. This reduces bias and can increase the trustworthiness of AI (Yarger et al., 2020). Thus, we will additionally assess how inquiring versus not inquiring about personal information during the AI-based selection process affects applicants’ trust in such AI-based selection processes. The aim of this study is therefore to investigate how the selected variables, personal information inquiries and feedback opportunities, affect applicants’ trust in AI systems in the context of hiring processes.

When actions yield predictable effects, we form bi-directional action-effect associations (e.g. Elsner & Hommel, 2001; Kunde, 2001). In turn, anticipating a desired effect allows us to select appropriate actions. Furthermore, Pfeuffer et al. (2016) discovered, that effect anticipation does not only affect effect-generating, for instance, manual responses but also eye movements. Participants anticipatorily look towards the future location of their actions‘ predictable effects (proactive effect monitoring; see also e.g., Gouret & Pfeuffer, 2021, preprint; Pfeuffer et al., 2022). These anticipatory saccades reflect the effect a person anticipates/expects and support a later comparison between expected and actual effect (e.g., Pfeuffer et al., 2016, 2022). In a prior study (Pfeuffer, 2023, https://osf.io/4wtvy), we investigated participants' win/loss expectations when betting points in a one-armed bandit game by assessing their eye movements during an anticipatory interval before feedback regarding their win/loss in a round was shown. Crucially, win and loss feedback consistently appeared at certain predictable locations. Our goal was to relate anticipatory eye movements towards the future win/loss feedback (i.e., corresponding expectations) to gambling-related behavior as well as attitudes/beliefs, and sensitivity to reward and punishment. This first study showed little evidence for relations between anticipatory eye movements and gambling-related behavior assessed via questionnaires. However, it had the crucial downside that we only offered a reward to the person who would reach the highest score. Thus, the task setting makes it very likely that participants might not believe they have an opportunity to win the corresponding reward and might thus not have given their best to win an individual reward and attended to feedback correspondingly. Thus, in this second experiment, we will replicate the one-armed bandit task with the crucial change that participants are informed that they can attain an individual reward based on their own performance in case they end the task with more points than 50% of the participants in the prior experiment. Gambling phase of the experiment (one-armed bandit game): In this first phase of the experiment, we will track participants’ anticipatory eye movements towards the future win/loss feedback to assess their win/loss expectations when betting points in a one-armed bandit game. They have the opportunity to win a reward based on the number of points they score. Importantly, note that wins and losses are randomly allocated both in the prior experiment and in this experiment. Thus, the actual point score is mainly based on luck (i.e., winning when more points are bet and losing when fewer points are bet). Exogenous attention phase of the experiment: Attention is affected by selection and reward history and corresponding biases (e.g., Awh et al., 2012). Importantly,reward is a source of such selection biases even several days after the initial association between stimulus and reward (e.g., Anderson et al., 2014; Della Libera & Chelazzi, 2009). To further extend our prior experiment, we will additionally assess differences in exogenous attention allocation between stimuli that signaled win/loss in the one-armed bandit game and a neutral stimulus. This will provide us with a means of additionally assessing reactive (i.e., non-anticipatory/non-proactive) saccades based on the presentation of previously win-/loss-related stimuli. We will additionally relate these reactive saccade patterns to proactive saccade patterns, gambling-related behavior as well as attitudes/beliefs, and sensitivity to reward and punishment. This will allow us to extend our perspective and to also investigate similarities and differences in proactive and reactive shifts of attention towards the respective stimuli. In the exogenous attention phase, we will assess the latency of participants’ saccades towards the win/loss and a neutral stimulus that unpredictably appear at the prior left/right effect/feedback locations. We expect participants to associate the win effect/feedback with reward (i.e., faster saccades) and for this to affect saccade latency. That is, we expect faster saccades towards the stimulus associated with a prior win as compared to the neutral stimulus. For the stimulus associated with loss (i.e., punishment), one could either expect it to be more salient than a neutral stimulus and thus associated with faster saccade latencies or one could expect that loss stimuli become associated with avoidance and an inhibitory process which would show in slower saccade latencies as compared to the neutral condition. Thus, we formulate an undirected hypothesis with regards to loss stimuli. Questionnaire phase: In a final phase, we will use questionnaires to assess gambling-related behavior as well as attitudes/beliefs, and sensitivity to reward and punishment and relate it to eye movements in the gambling phase and exogenous attention phase.

A person’s concerns about climate change do not necessarily translate into corresponding actions. In this context, it has been suggested that instilling a sense of urgency leads to more corresponding pro-environmental action (e.g., Brügger et al., 2015; Fesenfeld & Rinscheid, 2021). Moreover it is not only perceived urgency that appears to affect climate change-related beliefs, thoughts, or actions. Climate change is typically framed in and perceived in a temporal context (e.g., Ho, 2021; Kim & Ahn, 2019; Ma et al., 2022; McMichael & Katonivualiku, 2020; Pahl et al., 2014). That is, perceptions and actions related to climate change have been connected to temporal perception. Here, we build on two experiments which demonstrated severe underestimations of the time (number of years) between two pictures of the same location in the Alps (Pfeuffer & Heckmann, 2023, https://osf.io/jrgk4; Pfeuffer, Heckmann, & Hoffmann, 2023, https://osf.io/rkgsj). In a previous experiment (Pfeuffer, Heckmann, & Hoffmann, 2023), we used a framing manipulation to assess whether such temporal underestimations emerge due to how information regarding climate change is (publicly) communicated. Against our expectations, there was no main effect of framing, but we found interactions between framing (objective vs. climate change as war) and age (and between framing and climate change-related anxiety). That is, temporal underestimations of the time it took glaciers to melt decreased with age for the objective framing group but not for the war framing group. Here, we aim to ascertain the relationship between framing (objective vs. climate change as war) and age (young vs. old group). Participants are first confronted with one of two framing texts which either describe climate change objectively or as war. Then, participants are presented with pairs of pictures that show the same location in the Alps (recent vs. historic). The number of years between the two pictures of a pair varies and participants are asked to estimate how many years have passed between the recent and historic picture. Furthermore, we assess participants' pro-/anti-environmental attitudes via the New Ecological Paradigm Scale (NEP; Dunlap et al., 2000; Schleyer-Lindenmann et al., 2018) and their climate change related anxiety via the Climate Anxiety Scale (CAS; Clayton & Karazsia, 2010; Wullenkord et al., 2021).

This study assesses how recent as compared to historic pictures of the same location in the mountain glacial landscape of the Alps are evaluated (liking and wanting rating) and how this evaluation is affected by picture order (recent-historic vs. historic-recent) as well as individual pro-/anti-environment attitudes and climate change anxiety. A geographical research question is addressed with this study at the same time (see parallel preregistration).

With the rise of AI technologies, there are also rising customer concerns about data privacy, especially provoked by various security scandals and recent data leakages. Customer concerns are further intensified due to lacking reliable conveyance of the data security features a product does and does not entail. Such customer concerns mainly exist due to a relatively slow adoption of new technology (Diffusion of Innovation Theory; Rogers 2003). This is also related to the technology acceptance model (TAM; Davis, 1989; Marangunić & Granić, 2014) which suggests that perceived usefulness and perceived ease of use influence the final technology acceptance and, thus, actual usage of technology. Concerns of consumers in the face of new AI-based technology can also be explained with the concept of AI anxiety (Wang & Wang, 2019) that extends the well-recognised concept of computer anxiety (Zhang et al., 2021). Security labels such as the Digital Trust Label (Swiss Digital Initiative, 2022) that can be compared to food nutrition labels (e.g., NutriScore), could be used to achieve transparency regarding data security features of AI technologies in an easy and fast to process format. In turn, this could help achieve an appropriately low or high level of trust in corresponding technologies and resolve customers' concerns regarding the use and adoption of AI products. Such labels do not yet widely exist and it is yet unclear how they would need to be designed to support transparency and increase trust in consumers. Thus, assessing digital trust labels is an interesting area of research essential to support trust in AI products in the future. Prior studies suggest to convey the security of AI products through labels and certificates by third parties (Hammer et al., 2015; Wißner et al., 2014), as labels issued by third parties are perceived as more trustworthy and reliable (Josang et al., 2007). To date, security labels have only been studied to a small extent and, to our knowledge, no studies on security labels for AI products have been conducted yet. This is mostly due to the relatively new emergence of such labels and their current sparse application. For instance, Emami-Naeini et al. (2019) have conducted a qualitative study on a prototype Internet of Things (IoT) label (text and ratings) which found that information regarding security features is almost impossible to find for consumers and demonstrated that security and privacy the are most valued features after a product's price. Their label was shown to be accessible and useful to almost all study participants, encouraging them to consider the privacy and security in their future IoT buying decisions (Emami-Naeini et al., 2019). Furthermore, Shen and Vervier (2019) have designed two IoT labels based on previous research, one comparable to a food nutrition label including ticked boxes and colour, and one which is a table-like label with leveraged icons and text, that they both suggest for wider implementation given a decent mixture of complexity (amount of conveyed information) and simplicity. In this study, two different types of security labels (graphical vs. text-based labels) are compared, to understand the effect on consumers’ trust and AI anxiety as well as the value attributed to the AI product. The study aims to provide first suggestions regarding the design and usage of digital trust labels. Should such labels prove to be effective, concrete implications for businesses and governments could be derived that ultimately increase transparency and trust in AI products, creating a win-win situation for all parties involved.

Picture-word interference effects are interpreted as congruency effects comparable to Stroop effects (e.g., Piai et al., 2013; Starreveld & La Heij, 2017). Here, we investigate whether corresponding picture-word interference effects (i.e., congruency effects) also emerge for picture stimuli depicting climate change consequences (negative valence) versus unaffected nature (positive valence) and positive versus negative emotional words.

The aim of the study is to find out whether subjective landscape attractiveness (liking and wanting) of pictures of alpine landscapes in Kauner Valley (AT) and Martell Valley (IT) can be predicted based on intersubjective indices derived from digital elevation models (DEMs) and landcover maps. Furthermore, eye tracking will be used to explore attentional hot spots and gaze sequences across the images. Specifically, we will explore differences in how participants allocate their attention on historic and recent states of the landscape – as depicted on original and digitally rendered images of the same views – in two different alpine study areas (Kauner Valley and Martell Valley). A psychological research question is addressed with this study at the same time (see parallel preregistration, https://osf.io/9tnkp).

As the online learning market has also grown due to advancing digitalisation, AI solutions are coming to prominence here as well (Clark, 2020). In today's volatile, uncertain, complex and ambiguous (VUCA) environment, it is not only companies that need to constantly adapt. Employees, too, must continuously learn on their own initiative in order to remain employable (Hutasuhut et al., 2021; Joachims, 2022; Roux, 2021).The “process in which individuals take the initiative, with or without the help of others in diagnosing their learning needs, formulating learning goals, identifying human and material resources for learning, choosing and implementing appropriate learning strategies, and evaluating learning outcomes” (Knowles, 1975, p.18) is also referred to as Self-directed Learning (SDL). Studies have shown that, among other psychological phenomena, self-efficacy enables people to learn self-directedly (Long, 2007). Additionally, according to (Embo et al., 2010) SDL depends on feedback and support from supervisors. The use of artificial intelligence in e-learning systems can also support learners by providing personalized and tailored learning content (Murtaza et al., 2022). The present study aims to compare the employee evaluation of support from a human expert and an artificial intelligence in learning content selection in an e-learning environment. Here, participants will be presented with the scenario on a corporate e-learning platform, with either an expert or AI assisting them in putting together a personalized, custom-fit training plan. Half of the participants are randomly assigned to the expert (photographs, varying in ethnicity and gender, randomized, person only presumably assigned), the other half to the AI assistance condition. In a self-reflection task, participants provide information about selected skills, whether they are needed in their role, their own skill level, and a skill they would like to develop in the next 3 months. In both assistance conditions, participants will receive the training plan containing corresponding learning content on the e-learning platform after a matched delay (framed as the time frame for the expert to prepare the feedback or a corresponding delay to match conditions in the AI assistance condition). The display of the learning content is based on the information the participants provided regarding which skill they would like to develop in 3 months and which skill level they have in it. The study aims to find out how expert and AI support affects learner satisfaction, perceived effort, perceived use, technology acceptance, and learner autonomy. Furthermore, influences of occupational self-efficacy on these relationships are considered.

A person’s concerns about climate change do not necessarily translate into corresponding actions. In this context, it has been suggested that instilling a sense of urgency leads to more corresponding pro-environmental action (e.g., Brügger et al., 2015; Fesenfeld & Rinscheid, 2021). Moreover it is not only perceived urgency that appears to affect climate change-related beliefs, thoughts, or actions. Climate change is typically framed in and perceived in a temporal context (e.g., Ho, 2021; Kim & Ahn, 2019; Ma et al., 2022; McMichael & Katonivualiku, 2020; Pahl et al., 2014). That is, perceptions and actions related to climate change have been connected to temporal perception. Here, we aim at investigating the relationshipt between climate change anxiety and temporal estimations of changes in the mountain landscape of the Alps in the German population. Participants are presented with pairs of pictures that show the same location in the Alps with different durations of time in between the pictures of a pair and they are asked to estimate the number of years between the two pictures. Furthermore, we assess their climate change related anxiety via the Climate Anxiety Scale (CAS; Clayton & Karazsia, 2010; Wullenkord et al., 2021).

We will assess the justice perception of university students with respect to an algorithmic versus human decision-making processes regarding the distribution of topics (and presentation dates) in seminars (see Ackigoz et al., 2020; Lee, 2018, for relevant prior studies). The research question is as follows: Do justice perception, trust, and emotional responses of students differ between algorithmic versus human decision-making and does the communicated decisional fairness (information on decisional fairness vs. no information) affect this? The experimental design follows a 2x2 between-subjects design including the factors decision-making agent (human vs. algorithm) and communicated fairness (information on decisional fairness vs. no information). The outcome variables/dependent variables are perceived justice, trust, outrage, and satisfaction. Additionally, we will assess how outcome favourability affects the corresponding relationships.

When actions yield predictable effects, we form bi-directional action-effect associations (e.g. Elsner & Hommel, 2001; Kunde, 2001). In turn, anticipating a desired effect allows us to select appropriate actions. Furthermore, Pfeuffer et al. (2016) discovered, that effect anticipation does not only affect effect-generating, for instance, manual responses but also eye movements. Participants anticipatorily look towards the future location of their actions‘ predictable effects (proactive effect monitoring; see also e.g., Gouret & Pfeuffer, 2021, preprint; Pfeuffer et al., 2022). These anticipatory saccades reflect the effect a person anticipates/expects and support a later comparison between expected and actual effect (e.g., Pfeuffer et al., 2016, 2022). In the present study, I will assess participants' win/loss expectations when betting points in a one-armed bandit game by assessing their eye movements during an anticipatory interval before feedback regarding their win/loss in a round is shown. I will then relate these expectations (i.e., the corresponding eye movements) to gambling-related behavior as well as attitudes/beliefs, and sensitivity to reward and punishment.

When approaching a curve in the road, the area around the tangent point is vital for the driver to determine how to steer the car to stay on the road (e.g., Schnebelen et al. 2019). Prior studies on eye movements during driving suggest that an area around/close to the tangent point of a curve is fixated when approaching a curve both when actively driving towards a curve and when passively experiencing autonomous driving towards a curve (though differences are also apparent; see e.g., Mars & Navarro, 2012; Schnebelen et al., 2019). Moreover, Hergeth et al. (2016) showed that gaze behavior reflects the experienced trust/feeling of safety of drivers/passengers in an autonomous car. In a prior first experiment, we investigated the relationship between a driver’s perceived safety in an active or autonomous/passive driving situation (drive towards and through a curve) and the latency of the driver‘s first saccade into an area of interest around the tangent point of the curve as well as the fixation duration in that area of interest. We observed opposing relationships between saccade latency/fixation duration and perceived safety in active and autonomous/passive driving. To systematically further assess this relationship and determine whether safety perception affects gaze behavior or vice versa, in this experiment, we will use visual cues to guide participants‘ attention and gaze. We will compare three cueing conditions in active and autonomous/passive driving: No cueing, early cueing of the tangent point of a curve, and late cueing of the tangent point of a curve (with prior cueing of the road immediately ahead). As in our prior experiment, perceived safety is assessed via a corresponding rating after every driving situation.

Student self-assessment is an essential component of effective learning (Black & Wiliam, 1998) and accurate self-assessment (realistic subjective self-evaluations) can lead to more proactive learning (Karaman, 2021), arguably resulting in better exam performance. It is not clear, whether self-assessment has to be accurate to yield positive effects, since positively-biased self-assessment of students has also been found to have positive effects on performance, but inaccurate, negatively-biased self-assessment seems to be detrimental to performance (Butler, 2011). One way to measure how much self-evaluations align with actual performance (from now on denoted as self-evaluation accuracy as we will only focus on this aspect of self-assessment) is to compare the self-reported competence of a participant to the performance in a corresponding test (e.g., Narciss, Koerndle, & Dresel, 2011). However, in such studies explicit measures of self-evaluation are used, which are prone to self-report biases such as the tendency to be unrealistically optimistic about one`s abilities and the belief to be above average (Brown & Harris, 2014). Moreover, studies using such explicit measures of self-evaluation typically ask for global evaluations (e.g., course grade) rather than evaluations on the level of individual tasks or questions. Information on a learner’s self-evaluation and its accuracy could be beneficial to systematically guide the learning process, for instance, in e-learning by providing corresponding feedback. Yet frequent self-reports of subjective self-evaluation are not very practical and, as mentioned before, a subject to answer biases. Here, we will try to establish an alternative means of assessing self-evaluation (implicit and on the level of individual questions). Thus, we will test anticipatory saccades (e.g., Gouret & Pfeuffer, 2021; Pfeuffer et al., 2016, 2022) as a potential implicit measure of self-evaluations in the context of learning. Ideally, such anticipatory saccades could, in the future, be simultaneously assessed when learners complete, for instance, quizzes and should not be affected by answer biases. When an action contingently yields a consequence at a certain location (e.g., right key press > visual effect on the right), participants anticipatorily move their eyes in the direction of the future consequence/effect before it appears (e.g., Pfeuffer et al., 2016, 2022). Importantly, such anticipatory saccades already emerge after 1-2 learning trials (Gouret & Pfeuffer, 2021). We will test whether such anticipatory saccades also emerge when learning is tested in quizzes. Quiz questions will be presented in the central part of the screen and feedback will appear after a short delay. Crucially, correct feedback will appear on one side of the screen (e.g., left) and incorrect feedback will appear on the other side of the screen (e.g., right). We will assess whether participants’ anticipatory saccades (frequency of saccades in the direction of the future corresponding correct/incorrect feedback) can be used to assess participants’ expectations regarding the correctness of their respective responses and hence provide an implicit measure of their self-evaluation derived from the self-evaluation of individual answers. We will assess relations between implicit self-evaluation accuracy (I-SE-A; difference between anticipated and actual answer correctness across questions, sensitivity index according to signal detection theory) and implicit self-evaluation bias (I-SE-B; deviations of anticipated answer correctness from actual answer correctness in self-report questions in the positive or negative direction, response bias according to signal detection theory; both as inferred from such anticipatory saccades), quiz performance (multiple quizzes), accuracy and bias of self-reported explicit self-evaluations (E-SE-A and E-SE-B), and corresponding exam success. List of Abbreviations I-SE-A: Implicit Self-Evaluation Accuracy I-SE-B: Implicit Self-Evaluation Bias E-SE-A: Explicit Self-Evaluation Accuracy E-SE-B: Explicit Self-Evaluation Bias I-SE-A-X: Implicit Self-Evaluation Accuracy in Session X

In experiments on colour-word contingency learning, certain words are presented more/less frequently in certain colours (frequent vs. infrequent pairs - e.g., 80% vs. 20%) and thus require corresponding responses more/less frequently. Here, it is commonly observed that participants are faster to respond (and less error prone) for frequent as compared to infrequent colour-word pairs, showing that they learned the corresponding colour-word contingencies. In this study, we will assess whether colour-word contingencies also affect anticipatory eye movements (frequency and latency) to the predictable future consequences (coloured circles displayed on the left/right) of responses to colour-word pairs. That is, each left/right response to a colour-word pair will produce a 100% contingent effect (coloured circle) on the corresponding left/right side. This effect will appear after a short versus long blank screen interval (effect delay) following the response which is contingent with the frequent or infrequent combination. This is counterbalanced within participant where the first half has one pattern of frequent combinations following a short (long) effect delay and vice versa for the infrequent combination. We will use four word stimuli of which two will be paired with red/green and corresponding responses in 80% of trials and one of these word stimuli each will be paired with a short/long effect delay (100% contingency).

Humans are able to intentionally forget previously learned information (as often measured using the directed forgetting, short DF, paradigm, for an overview see Pastötter et al., 2017). In the item-method of DF (see MacLeod, 1998 for a review), for instance, participants are instructed to memorize stimuli that are presented to them sequentially. Each sequentially presented stimulus is directly followed by a memory cue informing participants to either forget or remember the stimulus. People typically recall fewer to-be-forgotten (TBF) than to-be-remembered (TBR) stimuli (Bjork, 1970). Whereas various studies have already demonstrated that humans are able to intentionally forget previously learned information, only few studies have analyzed whether DF affects motor representations (Tempel & Frings, 2016) or incidentally learned information (e.g., Hockley et al., 2016; Jou, 2010). Recent studies provided first evidence that typical findings regarding retrieval-induced forgetting and effects of DF extend to motor memory (in particular, motor sequences, Tempel et al., 2016; Tempel & Frings, 2013, 2016) and stimulus-response (S-R) associations (Dreisbach & Bäuml, 2014). Stimulus-response (S-R) associations are formed when stimuli and responses repeatedly co-occur and thus bind together – a notion that is supported by (item-specific) repetition priming effects (see Logan, 1988, 1990). Here, participants’ responses are, for instance, generally faster for stimuli that require the same as opposed to a different, previously-executed response. The current study aims to gain a deeper understanding of the mechanisms underlying the DF effect on the retrieval of S-R associations. For this, we will investigate whether (not explicitly instructed) learned S-R associations can be intentionally forgotten based on forget/remember cues following stimuli. To address this question, we will combine the DF item-method paradigm and an item-specific priming paradigm assessing the encoding and retrieval of S-R associations (see below and also Hsu & Waszak, 2012; Moutsopoulou et al., 2015). Information on the task and study design Our study will consist of three major phases: (1) A learning, (2) a distraction, and (3) a test phase. 1. Learning Phase: During the learning phase, participants’ task is to classify images of objects as containing a mechanism or not. We will extend the DF paradigm to S-R mappings using a classification task (drawing on an item-specific priming paradigm; Hsu & Waszak, 2012; Moutsopoulou et al., 2015; Pfeuffer, Moutsopoulou et al., 2017). By categorizing the stimuli, participants encode S-R associations (stimulus-action, S-A, associations between stimuli and motor outputs according to the terminology provided by Moutsopoulou et al., 2015). Importantly, whether a left or right response is required for the classification of a specific stimulus will differ between stimuli and will be indicated by a task cue detailing the classification-response mapping prior to the stimulus’ presentation. For example, the task cue “M + N” will indicate that an object image has to be classified as “mechanic” by pressing a left key and as “non-mechanic” by pressing a right key. Conversely, the task cue “N + M” will indicate that a right key press is required for mechanic objects and a left key press for non-mechanic ones. Participants will respond to each stimulus four times (in four blocks, see detailed task description below) in the exact same way (i.e., same S-R mapping). Per block, participants will also be instructed to memorize some of the object images for a later memory test and forget others. Directly after the response to a stimulus, a memory cue will inform participants to either forget or remember the stimulus. The same memory cue will appear after each of the four presentations of a specific stimulus. 2. Distraction Phase: In a subsequent distractor task, participants will solve a visual working memory task (Corsi) for 1.5 minutes to spurge short-term memory. 3. Test Phase: In the test phase, participants will be presented once more with the stimuli from the learning phase (both TBF and TBR objects) and new object images. Importantly, for the old objects, half of the stimuli per memory condition, the S-R mapping will be the same as in the learning phase (item-specific response repetition between learning and test). For the other half of the stimuli, the S-R mapping will be the opposite (item-specific response switch between learning and test). Response switches are therefore defined as item-specific switches of S-R mappings between the learning and the test phase of a stimulus (e.g., the stimulus requires a right key press during all four presentations in the learning phase, but a left key press during the test phase). Response repetitions are in turn defined as item-specific repetitions of the required response between the learning and the test phase (e.g., the stimulus requires a right key press during both the learning and test phase). Typically, RTs and error rates for item-specific response repetitions between learning phase and test phase trials are lower as compared to item-specific response switches, indicating that S-R associations have been formed during the learning phase and were retrieved in the test phase. We refer to these effects as item-specific S-R effects. Trial structure and task instructions will be equivalent to the last appearance of stimuli during the learning phase except that no memory cues will be presented anymore. Participants will be instructed that they will only need to respond to stimuli fast and accurately and that there will be no memory cues for this phase. Participants will also be informed that there will be no later memory test. They are told that they will respond to both TBR and TBF stimuli and that they should simply respond as fast and accurately as possible regardless of the previous memory instruction. 4. Distraction Phase: Participants will again solve a visual working memory task (Corsi) for 1.5 minutes to spurge short-term memory. How we aim to measure a DF effect on S-R associations We will use the difference in RTs and error rates in the test phase between trials that require an item-specific response switch and an item-specific response repetition to determine a DF effect based on the remember/forget cues. In other words, we will measure the strength of item-specific S-R effects for TBF as compared to TBR stimuli. Applying the observations in typical DF paradigms to the context of S-R associations we expect the following observations: If DF influences the retrieval of S-R associations, S-R effects, the performance differences between item-specific response repetitions and response switches in the test phase (see Hypotheses section) should be smaller for TBF as compared to TBR stimuli. Additionally, we will assess a DF effect on S-R associations by looking at the decrease in RTs and error rates across the four presentations of a stimulus (repetition priming effect) in the learning phase and compare them between TBR and TBR stimuli.

Every day we perform thousands of actions and most of these actions are followed by an effect. If an action is contingently followed by a specific effect, we form bi-directional action-effect associations (e.g., Elsner & Hommel, 2001; Kunde, 2001). This already happens after only a few pairings of action and effect (e.g., Gouret & Pfeuffer, 2021; Wolfensteller & Ruge, 2011). These bi-directional action-effect associations allow us to select appropriate actions based on desired/anticipated effects (e.g., Elsner & Hommel, 2001; Hommel et al., 2001; Kunde, 2001). Evidence for the impact of anticipated effects on action selection stems, for instance, from studies manipulating response-effect compatibility (e.g., Kunde, 2001). That is, responses to a target are faster when the response and the (predictably) following effect are (spatially) compatible (e.g., right button ➤ visual effect on the right) than when they are (spatially) incompatible (e.g., right button ➤ visual effect on the left) (response-effect, R-E, compatibility). Thus, there is evidence that the anticipation of our actions' future effects affects action selection (of effect-generating actions). Furthermore, we anticipatorily move our gaze towards the locations at which the (predictable) effects of our actions will appear in the future (e.g., Gouret & Pfeuffer, 2021; Pfeuffer et al., 2016). That is, there are more eye movements towards the future effect's location (effect-congruent) than in the opposite direction between the offset of a forced or free choice target indicating a manual response and the onset of effect (visual circle appearing) of this action. Moreover, the latency of participants‘ first anticipatory saccade towards a future effect (effect-congruent) depends upon the delay between action and effect (Gouret & Pfeuffer, preprint/2022). That is, the first effect-congruent saccade occurs later when the interval between action and effect is longer. According to Pfeuffer et al. (2016), such anticipatory eye movements reflect the proactive directing of attention towards the locations at which our actions will predictably cause an effect in the future. Pfeuffer et al. (2016) argued that such anticipatory saccades demonstrate a proactive effect monitoring process that enables a subsequent comparison of actual and expected effect. That is, the anticipation of our actions‘ future effects also triggers a proactive effect monitoring process (i.e., cognitive control). Moreover, our recent experiments demonstrated that action selectoin and proactive effect monitoring appear to be coordinated with one another. When the interval between target and effect was fixed and the latency of the manual, effect-generating action was manipulated, the first effect-congruent saccade was performed earlier when the manual response was delayed (wait condition, waiting for go signal) versus performed immediately upon target presentation (direct condition). This suggests that action selection of effect-generating manual actions (i.e., manual reaction times) and proactive effect monitoring (i.e., anticipatory saccades to future effect) are coordinated with one another – potentially, because they rely on the same cognitive resource – even though they are functionally dissociable processes (e.g., Pfeuffer et al., 2016). Finally, Bouhlel et al. (2022, unpublished experiment as part of the Psychology curriculum at the University of Freiburg) further explored the relationship between proactive effect monitoring and action selection when comparing active and passive manual responses (turning a steering wheel yourself vs. holding the wheel while a motor turns it). In this experiment, which used a more complex action than prior experiments on anticipatory saccades which relied on left/right button presses, anticipatory saccades were, on average delayed until after both direct (direct condition) and delayed (wait condition) manual responses. In this study, we will therefore assess the impact of action complexity on the occurrence and timing of anticipatory saccades by comparing manual responses of different complexity levels (key press to turning a steering wheel).

Humans are able to intentionally forget previously learned information (as often measured using the directed forgetting, short DF, paradigm, for an overview see Pastötter, Tempel, & Bäuml, 2017): In the item-method of DF (see MacLeod, 1998 for a review), for instance, participants are instructed to memorize stimuli that are presented to them sequentially. Each sequentially presented stimulus is directly followed by a cue informing participants to either forget or remember the stimulus. People typically recall fewer to-be-forgotten (TBF) than to-be-remembered (TBR) stimuli (Bjork, 1970). Whereas various studies have already demonstrated that humans are able to intentionally forget previously learned information, only few studies have analyzed whether DF affects motor representations (Tempel & Frings, 2016) or incidentally learned information (e.g., Hockley, Ahmad, & Nicholson, 2016; Jou, 2010). Recent studies provided first evidence that typical findings regarding retrieval-induced forgetting and effects of DF extend to motor memory (in particular, motor sequences, Tempel, Aslan, & Frings, 2016; Tempel & Frings, 2013, 2016) and stimulus-response (S-R) associations (Dreisbach & Bäuml, 2014). Stimulus-response (S-R) associations are formed when stimuli and responses repeatedly co-occur and thus bind together – a notion that is supported by (item-specific) repetition priming effects (see Logan, 1988, 1990). Here, participants’ responses are, for instance, generally faster for stimuli that require the same as opposed to a different, previously-executed response. The current study aims to gain a deeper understanding of the mechanisms underlying the DF effect on the retrieval of S-R associations. For this, we will investigate whether (not explicitly instructed) learned S-R associations can be intentionally forgotten based on forget/remember cues following stimuli. To address this question, we will combine the DF item-method paradigm and an item-specific priming paradigm assessing the encoding and retrieval of S-R associations (see below and also Hsu & Waszak, 2012; Moutsopoulou, Yang, Desantis, & Waszak., 2015). ***Information on the task and study design*** Our study will consist of five major phases: (1) A learning, (2) a distraction, (3) a test, (4) another distraction, and (5) a recognition phase. 1. Learning Phase: During the learning phase, participants’ task is to classify images of objects as containing a mechanism or not. We will extend the DF paradigm to S-R mappings using a classification task (drawing on an item-specific priming paradigm; Hsu & Waszak, 2012; Moutsopoulou, Yang, Desantis, & Waszak, 2015; Pfeuffer, Moutsopoulou, Pfister, Waszak, & Kiesel, 2017). By categorizing the stimuli, participants encode S-R associations (stimulus-action, S-A, associations between stimuli and motor outputs according to the terminology provided by Moutsopoulou et al., 2015). Importantly, whether a left or right response is required for the classification of a specific stimulus will differ between stimuli and will be indicated by a task-cue detailing the classification-response mapping prior to the stimulus’ presentation. For example, the task cue “M + N” will indicate that an object image has to be classified as “mechanic” by pressing a left key and as “non-mechanic” by pressing a right key. Conversely, the task cue “N + M” will indicate that a right key press is required for mechanic objects and a left key press for non-mechanic ones. Participants will respond to each stimulus four times (in four blocks, see detailed task description below) in the exact same way (i.e., same S-R mapping). In the fourth block, participants will also be instructed to memorize some of the object images for a later memory test. In this block, directly after the response to a stimulus, a cue will inform participants to either forget or remember the stimulus. 2. Distraction Phase: In a subsequent distractor task, participants will solve a visual working memory task for 1.5 minutes to spurge short-term memory. 3. Test Phase: In the test phase, participants will be presented once with the stimuli from the learning phase (both TBF and TBR objects). Importantly, for half of the stimuli, the S-R mapping are the same as in the learning phase (item-specific response repetition between learning and test). For the other half of the stimuli, the S-R mapping will be the opposite (item-specific response switch between learning and test). Response switches are therefore defined as item-specific switches of the stimuli´ S-R mappings between the learning and the test phase (e.g., stimulus requires a right key press during the learning phase, but a left key press during the test phase). Response repetitions are in turn defined as item-specific repetitions of the required response between the learning and the test phase (stimulus requires a right key press during both the learning and test phase). Typically, RTs and error rates for item-specific response repetitions between learning phase and test phase trials are lower as compared to item-specific response switches, indicating that S-R associations have been formed during the learning phase and were retrieved in the test phase. We refer to these effects as item-specific S-R effects. Trial structure and task instructions will be equivalent to the last appearance of stimuli during the learning phase. Again, after responding to each stimulus, the same remember or forget cue as in the 4th block of the learning phase will be presented. Participants will be reminded to only remember the TBR stimuli. 4. Distraction Phase: Participants will again solve a visual working memory task for 1.5 minutes to spurge short-term memory. 5. Recognition Phase: In the recognition phase, participants will be presented with all the object images from the learning phase and an equal number of new object images. Participants will be instructed to identify whether an object has previously been presented (category: “old”) or not (category: “new”) – regardless of its associated memory instruction – by pressing either left or right response key. On a given trial, an old or new stimulus image is presented in the center of the screen, and participants are instructed to categorize the stimulus as quickly and accurately as possible as either “old” or “new” according to the category labels that are displayed in the right or left upper screen corner. ***How we aim to measure a DF effect on S-R associations*** We will use the difference in RTs and error rates between trials that require an item-specific response switch and an item-specific response repetition to determine a DF effect based on the remember/forget cues. In other words, we will measure the strength of item-specific S-R effects for TBF as compared to TBR stimuli. Applying the observations in typical DF paradigms to the context of S-R associations we expect the following observations: If DF influences the retrieval of S-R associations, S-R effects, the performance differences between item-specific response repetitions and response switches (see Hypotheses section) should be smaller for TBF as compared to TBR stimuli.

When actions yield predictable effects, we form bi-directional action-effect associations (e.g. Elsner & Hommel, 2001; Kunde, 2001). In turn, anticipating a desired effect allows us to select appropriate actions . That is, anticipated effects of our actions influence the selection of effect-generating actions. For instance, participants showed significantly shorter reaction times (RTs), when their manual reaction reliably yielded a spatially compatible (visual) effect (e.g. left key press > effect on left side of the display) as opposed to a spatially incompatible effects (e.g. left key press > effect on right side of the display). Furthermore, Pfeuffer et al. (2016) discovered, that effect anticipation does not only affect effect-generating, for instance, manual responses but also eye movements. Participants look significantly more often towards the side of the display on which the future effect of their action will appear (effect-congruent) rather than towards the opposite side(saccade-effect congruency,SEC, effect; see also e.g., Gouret & Pfeuffer, 2021). That is, effect anticipation also triggers anticipatory saccades towards future effects. Pfeuffer et al. (2016) hypothesized that these anticipatory saccades serve the function of proactive effect monitoring. That is, we anticipatorily saccade towards the future location of our effects to aid the subsequent comparison of expected and actual effect. Furthermore, Gouret and Pfeuffer (preprint) discovered that not only the location but also the timing of future effects are anticipated and proactively monitored. In their experiment, left/right key presses were associated with a short versus long effect delay (e.g., left key press > short delay; right key press > long delay). They found that the latency of the first saccade towards the future effect was longer when the effect appeared after a long rather than short effect delay. Importantly, in prior experiments, effects were typically irrelevant for the task at hand (and also not associated with rewards etc.). Thus, there was little reason to proactively monitor them and perform anticipatory saccades. Nevertheless, prior studies observed a substantial number of anticipatory saccades. Here, we will further investigate the impact of effect relevance on proactive effect monitoring (i.e., corresponding anticipatory saccades). That is, we will manipulate the timing of relevant and irrelevant features of future effects and assess whether proactive effect monitoring (i.e. anticipatory saccades) is directed towards every, respectively the first appearing effect feature, regardless of its relevance for the task or whether proactive effect monitoring is specifically aimed at relevant effect features (when effects contain both relevant and irrelevant features).

Humans are able to intentionally forget previously learned information (as often measured using the directed forgetting, short DF, paradigm, for an overview see Pastötter et al., 2017). In the item-method of DF (see MacLeod, 1998 for a review), for instance, participants are instructed to memorize stimuli that are presented to them sequentially. Each sequentially presented stimulus is directly followed by a memory cue informing participants to either forget or remember the stimulus. People typically recall fewer to-be-forgotten (TBF) than to-be-remembered (TBR) stimuli (Bjork, 1970). Whereas various studies have already demonstrated that humans are able to intentionally forget previously learned information, only few studies have analyzed whether DF affects motor representations (Tempel & Frings, 2016) or incidentally learned information (e.g., Hockley et al., 2016; Jou, 2010). Recent studies provided first evidence that typical findings regarding retrieval-induced forgetting and effects of DF extend to motor memory (in particular, motor sequences, Tempel et al., 2016; Tempel & Frings, 2013, 2016) and stimulus-response (S-R) associations (Dreisbach & Bäuml, 2014). Stimulus-response (S-R) associations are formed when stimuli and responses repeatedly co-occur and thus bind together – a notion that is supported by (item-specific) repetition priming effects (see Logan, 1988, 1990). Here, participants’ responses are, for instance, generally faster for stimuli that require the same as opposed to a different, previously-executed response. The current study aims to gain a deeper understanding of the mechanisms underlying the DF effect on the retrieval of S-R associations. For this, we will investigate whether (not explicitly instructed) learned S-R associations can be intentionally forgotten based on forget/remember cues following stimuli. To address this question, we will combine the DF item-method paradigm and an item-specific priming paradigm assessing the encoding and retrieval of S-R associations (see below and also Hsu & Waszak, 2012; Moutsopoulou et al., 2015). Information on the task and study design Our study will consist of three major phases: (1) A learning, (2) a distraction, and (3) a test phase. 1. Learning Phase: We will extend the DF paradigm to S-R mappings using a classification task (drawing on an item-specific priming paradigm; Hsu & Waszak, 2012; Moutsopoulou et al., 2015; Pfeuffer, Moutsopoulou et al., 2017). During the learning phase, participants’ task is to classify words (referring to concrete objects) as animate or inanimate (animacy task). By categorizing the stimuli, participants encode S-R associations (stimulus-action, S-A, associations between stimuli and motor outputs according to the terminology provided by Moutsopoulou et al., 2015). Importantly, whether a left or right response is required for an animate/inanimate classification of a specific stimulus will differ between stimuli and will be indicated by a task cue detailing the classification-response mapping prior to the stimulus’ presentation. For example, the task cue “A + I” will indicate that a word has to be classified as “animate” by pressing a left key and as “inanimate” by pressing a right key. Conversely, the task cue “I + A” will indicate that a right key press is required for words denoting animate stimuli and a left key press for words denoting inanimate stimuli. At the same time, participants will be instructed to memorize some of the words for a later memory test and forget others. Directly after the response to a stimulus, a memory cue will inform participants to either forget or remember the stimulus. 2. Distraction Phase: In a subsequent distractor task, participants will solve a visual working memory task (Corsi) for 1.5 minutes to spurge short-term memory. 3. Test Phase: In the test phase, participants will be presented once more with the stimuli from the learning phase (both TBF and TBR words) and new words. Importantly, for the old words, half of the stimuli per memory condition, the S-R mapping will be the same as in the learning phase (item-specific response repetition between learning and test). For the other half of the stimuli, the S-R mapping will be the opposite (item-specific response switch between learning and test). Response switches are therefore defined as item-specific switches of S-R mappings between the learning and the test phase of a stimulus (e.g., the stimulus requires a right key press in the learning phase, but a left key press in the test phase). Response repetitions are, in turn, defined as item-specific repetitions of the required response between the learning and the test phase (e.g., the stimulus requires a right key press in both the learning and test phase). Typically, RTs and error rates for item-specific response repetitions between learning phase and test trials are lower as compared to item-specific response switches, indicating that S-R associations have been formed during the learning phase and were retrieved in the test phase. We refer to this effect as item-specific S-R effect. Trial structure and task instructions in the test phase will be equivalent to the learning phase except that no memory cues will be presented anymore. Participants will be instructed that they will only need to respond to stimuli fast and accurately and that there will be no memory cues for this phase. Participants will also be informed that there will be no later memory test. They are told that they will respond to both TBR and TBF stimuli and that they should simply respond as fast and accurately as possible regardless of the previous memory instruction. How we aim to measure a DF effect on S-R associations We will use the difference in RTs and error rates in the test phase between trials that require an item-specific response switch and an item-specific response repetition (S-R effect) to determine a DF effect based on the remember/forget cues. In other words, we will measure the strength of item-specific S-R effects for TBF as compared to TBR stimuli. If there was a DF effect on the retrieval or encoding of S-R associations, the performance differences between item-specific response repetitions and response switches (S-R effect) in the test phase (see Hypotheses section) should be smaller for TBF as compared to TBR stimuli.

item-specific priming paradigm assessing execution-based and verbal code-based S-C and S-A associations - the proportion of item-specific switches in S-C and S-A mapping is varied between participants