Background According to the seductive-details effect, practitioners should avoid interesting but irrelevant adjuncts (e.g., fun facts, comics) in learning materials as they might increase extraneous cognitive load and thus hamper learning. As the digitalisation of learning continues to increase, the question arises as to whether this recommendation also applies to interesting but irrelevant hyperlinks that are often included in online materials. Objectives We investigated whether students perceive and use seductive details that appear behind hyperlinks differently compared to seductive details that are integrated in the learning material directly on the screen. We were assuming that students would (a) perceive hyperlink seductive details as less relevant than integrated seductive details and thus not be negatively affected in their cognitive processing (i.e., informed use of seductive details) and (b) use hyperlink seductive details to take a small break or for gratification, thereby supporting them in their persistent online learning (i.e., needs-oriented use of seductive details). Methods In a 3 & times; 2-between-subjects study, participants (N = 165) worked online on a learning unit about chemistry models without seductive details or with them, either with explicit information about their irrelevance for the learning goal or without such explicit information. Moreover, we presented the details integrated in the learning material or as mouse-over hyperlinks. Results and Conclusion Although students perceived hyperlink seductive details as less relevant than integrated details, their learning outcomes were still impaired when they did not receive an additional explicit irrelevance instruction. Hence, our study reveals no evidence that students use seductive details via hyperlinks differently from integrated ones, but instead that seductive details via hyperlinks should also be avoided.
This work presents a characterization method for cameras with trichromatic RGB color filter array and polarization layer (RGB-P) sensor-based imaging devices. Such sensors enable the reconstruction of color and polarization of registered scene elements, which is an important requirement in computer vision. We will present spectral responsivity measurements, which reveal different sensitivities for various color and polarization channels. Furthermore, we will discuss and model an observed chromaticity shift in registered camera signals for polarized irradiance. Both lead to inaccurate estimation of color and polarization features. In order to overcome these issues, we will present a neural-network-based model for color and polarization feature reconstruction. Essentially, it considers spectral sensitivity for polarized irradiance. Furthermore, the model takes into account that, for visualization, the color signals have to be a linear combination of polarization channels. Models were trained for selected natural and synthetic reflectance sets, as well as commonly used lighting. We evaluated the resulting performance, which yielded robust results. The method can be employed for an estimation of color and polarization features for RGB-P imaging devices. Applications can be found in photography, as well as machine and computer vision, in which object surface color rendering plays a major role.
A Large Language Model (LLM) as judge evaluates the quality of victim Machine Learning (ML) models, specifically LLMs, by analyzing their outputs. An LLM as judge is the combination of one model and one specifically engineered judge prompt that contains the criteria for the analysis. The resulting automation of the analysis scales up the complex evaluation of the victim models' free-form text outputs by faster and more consistent judgments compared to human reviewers. Thus, quality and security assessments of LLMs can cover a wide range of the victim models' use cases. Being a comparably new technique, LLMs as judges lack a thorough investigation for their reliability and agreement to human judgment. Our work evaluates the applicability of LLMs as automated quality assessors of victim LLMs. We test the efficacy of 37 differently sized conversational LLMs in combination with 5 different judge prompts, the concept of a second-level judge, and 5 models fine-tuned for the task as assessors. As assessment objective, we curate datasets for eight different categories of judgment tasks and the corresponding ground-truth labels based on human assessments. Our empirical results show a high correlation of LLMs as judges with human assessments, when combined with a suitable prompt, in particular for GPT-4o, several open-source models with ⩾ 32B parameters, and a few smaller models like Qwen2.5 14B.
Stanley Milgram's ‘Obedience to Authority’ study demonstrated that an experimenter can lead participants to obey orders that seemingly causes others harm. To examine the nature of the experimenter's influence, we analysed the experimenter's non-procedural interventions in 136 sessions, across four experimental conditions. Results showed that, with a sole exception, Milgram's experimenter used each of his non-procedural interventions in less than half of both fully obedient sessions and the obedient phases of disobedient sessions. In stark contrast, the intervention ‘Please continue’ was used in virtually all obedient sessions as well as the obedient phases of disobedient sessions, and with substantial frequency. Indeed, results also demonstrated that had Milgram defined a session ‘disobedient’ when any standard prod other than ‘Please continue’ is required, the obedience rates obtained would have already been historic. These findings challenge both the neglect of the experimenter's interactional involvement in Milgram's original account and the focus on explicit acts of argumentation, manipulation, order or the forging of common identity in more contemporary research. Instead, the findings are congruent with the hypothesis that the experimenter operated implicitly, via minimal interpersonal gestures. The paper ends by discussing the claim that, belying their minimal content, these gestures may have contributed to obedience by symbolically deleting participants as active agents, and hence, coercing them.
Information seeking as a behavioral manifestation of curiosity has often been captured with experimental tasks that explicitly prompt information seeking instead of capturing participants' self-initiated exploratory behaviors. This online study (N = 799 English-speaking adults, mostly from the United Kingdom) was conducted in 2023 and combined aspects of experimental curiosity research with more naturalistic exploration methods by introducing a novel experimental setup. Participants freely explored a hypertext on a historical topic, while their self-directed information seeking (i.e., clicks on hyperlinks and reading time) was captured with log data. Participants then completed a knowledge test in which they also reported their confidence in their answers and their curiosity about learning the correct answers. We found that simply choosing to seek additional information (clicking on a hyperlink) did not predict curiosity ratings, whereas the extent of engagement (reading time) did. Moreover, trait curiosity moderated the relationship between confidence and state curiosity, with highly trait-curious individuals maintaining higher state curiosity under low confidence, unlike those with low trait curiosity. This study presents a promising way to assess curiosity in a controlled experimental task resembling real-world learning situations and provides a comprehensive perspective on the dynamic ways in which curiosity is sparked and satisfied. (PsycInfo Database Record (c) 2026 APA, all rights reserved).