Composing tailored cover letters for specific job applications is challenging, particularly when applying for multiple jobs. We present DUPLet, an interactive system that leverages large language models (LLMs) to produce more personalized and authentic cover letters. The key novelty of our system lies in dynamic user modeling that leverages a natural-language description of the user’s profile and preferences, which adapts over time based on provided input and interactions with the system. Based on a resume, job description, and a series of LLM-generated, job-specific questions, DUPLet creates an initial cover letter draft that users can then iteratively refine. We conducted a preliminary evaluation of DUPLet by simulating user interactions with LLMs and using the LLM-as-a-judge paradigm to rate the cover letters on dimensions, such as personalization and authenticity. Results indicate that cover letter variants that involve the user and leverage a dynamically refined natural-language user profile tend to receive higher ratings.

Video-based gaze estimation methods aim to capture the inherently temporal dynamics of human eye gaze from multiple image frames. However, since models must capture both spatial and temporal relationships, performance is limited by the feature representations within a frame but also between multiple frames. We propose the Spatio-Temporal Gaze Network (ST-Gaze), a model that combines a CNN backbone with dedicated channel attention and self-attention modules to fuse eye and face features optimally. The fused features are then treated as a spatial sequence, allowing for the capture of an intra-frame context, which is then propagated through time to model inter-frame dynamics. We evaluated our method on the EVE dataset and show that ST-Gaze achieves state-of-the-art performance both with and without person-specific adaptation. Additionally, our ablation study provides further insights into the model performance, showing that preserving and modelling intra-frame spatial context with our spatio-temporal recurrence is fundamentally superior to premature spatial pooling. As such, our results pave the way towards more robust video-based gaze estimation using commonly available cameras.

Modelling human gaze behaviour on 360 \({}^{\circ}\) images is important for various human–computer interaction applications. However, existing methods are limited to predicting discrete fixation sequences or aggregated saliency maps, thereby neglecting fine-grained gaze behaviour such as saccadic eye movements that can be captured by commercial eye-trackers. We introduce a more challenging task— fine-grained gaze sequence generation . This task aims to generate eye-tracker-like gaze data for given stimuli. We propose DiffGaze , a diffusion-based method for generating realistic and diverse fine-grained human gaze sequences conditioned on 360 \({}^{\circ}\) images. We evaluate DiffGaze on two 360 \({}^{\circ}\) image benchmarks for fine-grained gaze sequence generation as well as two downstream tasks, scanpath prediction and saliency prediction. Our evaluations show that DiffGaze outperforms the fine-grained gaze generation baselines in all tasks on both benchmarks. We also report a 21-participant survey study showing that our method generates gaze sequences that are indistinguishable from real human sequences. Taken together, our evaluations not only demonstrate the effectiveness of DiffGaze but also point towards a new generation of methods that faithfully model the rich spatial and temporal nature of natural human gaze behaviour.

Abstract While deep reinforcement learning (RL) agents outperform humans on an increasing number of tasks, training them requires data equivalent to decades of human gameplay. Recent hierarchical RL methods have increased sample efficiency by incorporating information inherent to the structure of the decision problem but at the cost of having to discover or use human-annotated sub-goals that guide the learning process. We show that intentions of human players, i.e. the precursor of goal-oriented decisions, can be robustly predicted from eye gaze even for the long-horizon sparse rewards task of Montezuma’s Revenge–one of the most challenging RL tasks in the Atari2600 game suite. We propose Int-HRL: Hierarchical RL with intention-based sub-goals that are inferred from human eye gaze. Our novel sub-goal extraction pipeline is fully automatic and replaces the need for manual sub-goal annotation by human experts. Our evaluations show that replacing hand-crafted sub-goals with automatically extracted intentions leads to an HRL agent that is significantly more sample efficient than previous methods.

We present HAIFAI—a novel two-stage system where humans and AI interact to tackle the challenging task of reconstructing a visual representation of a face that exists only in a person’s mind. In the first stage, users iteratively rank images our reconstruction system presents based on their resemblance to a mental image. These rankings, in turn, allow the system to extract relevant image features, fuse them into a unified feature vector and use a generative model to produce an initial reconstruction of the mental image. The second stage leverages an existing face editing method, allowing users to manually refine and further improve this reconstruction using an easy-to-use slider interface for face shape manipulation. To avoid the need for tedious human data collection for training the reconstruction system, we introduce a computational user model of human ranking behaviour. For this, we collected a small face ranking dataset through an online crowd-sourcing study containing data from 275 participants. We evaluate HAIFAI and an ablated version in a 12-participant user study and demonstrate that our approach outperforms the previous state of the art regarding reconstruction quality, usability, perceived workload and reconstruction speed. We further validate the reconstructions in a subsequent face ranking study with 18 participants and show that HAIFAI achieves a new state-of-the-art identification rate of 60.6%. These findings represent a significant advancement towards developing new interactive intelligent systems capable of reliably and effortlessly reconstructing a user’s mental image.

Understanding the link between visual attention and users’ information needs when visually exploring information visualisations is under-explored due to a lack of large and diverse datasets to facilitate these analyses. To fill this gap we introduce SalChartQA – a novel crowd-sourced dataset that uses the BubbleView interface to track user attention and a question-answering (QA) paradigm to induce different information needs in users. SalChartQA contains 74,340 answers to 6,000 questions on 3,000 visualisations. Informed by our analyses demonstrating the close correlation between information needs and visual saliency, we propose the first computational method to predict question-driven saliency on visualisations. Our method outperforms state-of-the-art saliency models for several metrics, such as the correlation coefficient and the Kullback-Leibler divergence. These results show the importance of information needs for shaping attentive behaviour and pave the way for new applications, such as task-driven optimisation of visualisations or explainable AI in chart question-answering.

The mouse is a pervasive input device used for a wide range of interactive applications. However, computational modelling of mouse behaviour typically requires time-consuming design and extraction of handcrafted features, or approaches that are application-specific. We instead propose Mouse2Vec – a novel self-supervised method designed to learn semantic representations of mouse behaviour that are reusable across users and applications. Mouse2Vec uses a Transformer-based encoder-decoder architecture, which is specifically geared for mouse data: During pretraining, the encoder learns an embedding of input mouse trajectories while the decoder reconstructs the input and simultaneously detects mouse click events. We show that the representations learned by our method can identify interpretable mouse behaviour clusters and retrieve similar mouse trajectories. We also demonstrate on three sample downstream tasks that the representations can be practically used to augment mouse data for training supervised methods and serve as an effective feature extractor.

While visual saliency has recently been studied in 3D, the experimental setup for collecting 3D saliency data can be expensive and cumbersome. To address this challenge, we propose a novel experimental design that utilises an eye tracker on a screen to collect 3D saliency data, which could reduce the cost and complexity of data collection. We first collected gaze data on a computer screen and then mapped the 2D points to 3D saliency data through perspective transformation. Using this method, we propose Saliency3D, a 3D saliency dataset (49,276 fixations) comprising 10 participants looking at sixteen objects. We examined the viewing preferences for objects and our results indicate potential preferred viewing directions and a correlation between salient features and the variation in viewing directions.

Reusable embeddings of user behaviour have shown significant performance improvements for the personalised saliency prediction task. However, prior works require explicit user characteristics and preferences as input, which are often difficult to obtain. We present a novel method to extract user embeddings from pairs of natural images and corresponding saliency maps generated from a small amount of user-specific eye tracking data. At the core of our method is a Siamese convolutional neural encoder that learns the user embeddings by contrasting the image and personal saliency map pairs of different users. Evaluations on two public saliency datasets show that the generated embeddings have high discriminative power, are effective at refining universal saliency maps to the individual users, and generalise well across users and images. Finally, based on our model's ability to encode individual user characteristics, our work points towards other applications that can benefit from reusable embeddings of gaze behaviour.

Question answering has recently been proposed as a promising means to assess the recallability of information visualisations. However, prior works are yet to study the link between visually encoding a visualisation in memory and recall performance. To fill this gap, we propose VisRecall++ -- a novel 40-participant recallability dataset that contains gaze data on 200 visualisations and 1,000 questions, including identifying the title and retrieving values. We measured recallability by asking participants questions after they observed the visualisation for 10 seconds. Our analyses reveal several insights, such as saccade amplitude, number of fixations, and fixation duration significantly differ between high and low recallability groups. Finally, we propose GazeRecallNet -- a novel computational method to predict recallability from gaze behaviour that outperforms the state-of-the-art model RecallNet and three other baselines on this task. Taken together, our results shed light on assessing recallability from gaze behaviour and inform future work on recallability-based visualisation optimisation.

We propose Unified Model of Saliency and Scanpaths (UMSS)- a model that learns to predict multi-duration saliency and scanpaths (i.e. sequences of eye fixations) on information visualisations. Although scanpaths provide rich information about the importance of different visualisation elements during the visual exploration process, prior work has been limited to predicting aggregated attention statistics, such as visual saliency. We present in-depth analyses of gaze behaviour for different information visualisation elements (e.g. Title, Label, Data) on the popular MASSVIS dataset. We show that while, overall, gaze patterns are surprisingly consistent across visualisations and viewers, there are also structural differences in gaze dynamics for different elements. Informed by our analyses, UMSS first predicts multi-duration element-level saliency maps, then probabilistically samples scanpaths from them. Extensive experiments on MASSVIS show that our method consistently outperforms state-of-the-art methods with respect to several, widely used scanpath and saliency evaluation metrics. Our method achieves a relative improvement in sequence score of 11.5% for scanpath prediction, and a relative improvement in Pearson correlation coefficient of up to 23.6% for saliency prediction. These results are auspicious and point towards richer user models and simulations of visual attention on visualisations without the need for any eye tracking equipment.

We introduce SUPREYES – a novel self-supervised method to increase the spatio-temporal resolution of gaze data recorded using low(er)-resolution eye trackers. Despite continuing advances in eye tracking technology, the vast majority of current eye trackers – particularly mobile ones and those integrated into mobile devices – suffer from low-resolution gaze data, thus fundamentally limiting their practical usefulness. SUPREYES learns a continuous implicit neural representation from low-resolution gaze data to up-sample the gaze data to arbitrary resolutions. We compare our method with commonly used interpolation methods on arbitrary scale super-resolution and demonstrate that SUPREYES outperforms these baselines by a significant margin. We also test on the sample downstream task of gaze-based user identification and show that our method improves the performance of original low-resolution gaze data and outperforms other baselines. These results are promising as they open up a new direction for increasing eye tracking fidelity as well as enabling new gaze-based applications without the need for new eye tracking equipment.

We introduce an end-to-end interactive system for mental face reconstruction – the challenging task of visually reconstructing a face image a person only has in their mind. In contrast to existing methods that suffer from low usability and high mental load, our approach only requires the user to rank images over multiple iterations according to the perceived similarity with their mental image. Based on these rankings, our mental face reconstruction system extracts image features in each iteration, combines them into a joint feature vector, and then uses a generative model to visually reconstruct the mental image. To avoid the need for collecting large amounts of human training data, we further propose a computational user model that can simulate human ranking behaviour using data from an online crowd-sourcing study (N=215). Results from a 12-participant user study show that our method can reconstruct mental images that are visually similar to existing approaches but has significantly higher usability, lower perceived workload, and is faster. In addition, results from a third 22-participant lineup study in which we validated our reconstructions on a face ranking task show a identification rate of , which is in line with prior work. These results represent an important step towards new interactive intelligent systems that can robustly and effortlessly reconstruct a user’s mental image.

Desktop notifications should be noticeable but are also subject to a number of design choices, e.g. concerning their size, placement, or opacity. It is currently unknown, however, how these choices interact with the desktop background and their influence on noticeability. To address this limitation, we introduce a software tool to automatically synthesize realistically looking desktop images for major operating systems and applications. Using these images, we present a user study (N=34) to investigate the noticeability of notifications during a primary task. We are first to show that visual importance of the background at the notification location significantly impacts whether users detect notifications. We analyse the utility of visual importance to compensate for suboptimal design choices with respect to noticeability, e.g. small notification size. Finally, we introduce noticeability maps - 2D maps encoding the predicted noticeability across the desktop and inform designers how to trade-off notification design and noticeability.

We propose Neuro-Symbolic Visual Dialog (NSVD) —the first method to combine deep learning and symbolic program execution for multi-round visually-grounded reasoning. NSVD significantly outperforms existing purely-connectionist methods on two key challenges inherent to visual dialog: long-distance co-reference resolution as well as vanishing question-answering performance. We demonstrate the latter by proposing a more realistic and stricter evaluation scheme in which we use predicted answers for the full dialog history when calculating accuracy. We describe two variants of our model and show that using this new scheme, our best model achieves an accuracy of 99.72% on CLEVR-Dialog—a relative improvement of more than 10% over the state of the art—while only requiring a fraction of training data. Moreover, we demonstrate that our neuro-symbolic models have a higher mean first failure round, are more robust against incomplete dialog histories, and generalise better not only to dialogs that are up to three times longer than those seen during training but also to unseen question types and scenes.

Gaze-based analysis of areas of interest (AOIs) is widely used in information visualisation research to understand how people explore visualisations or assess the quality of visualisations concerning key characteristics such as memorability. However, nearby AOIs in visualisations amplify the uncertainty caused by the gaze estimation error, which strongly influences the mapping between gaze samples or fixations and different AOIs. We contribute a novel investigation into gaze uncertainty and quantify its impact on AOI-based analysis on visualisations using two novel metrics: the Flipping Candidate Rate (FCR) and Hit Any AOI Rate (HAAR). Our analysis of 40 real-world visualisations, including human gaze and AOI annotations, shows that gaze uncertainty frequently and significantly impacts the analysis conducted in AOI-based studies. Moreover, we analysed four visualisation types and found that bar and scatter plots are usually designed in a way that causes more uncertainty than line and pie plots in gaze-based analysis.

One approach to mitigate shoulder surfing attacks on mobile devices is to detect the presence of a bystander using the phone’s front-facing camera. However, a person’s face in the camera’s field of view does not always indicate an attack. To overcome this limitation, in a novel data collection study (N=16), we analysed the influence of three viewing angles and four distances on the success of shoulder surfing attacks. In contrast to prior works that mainly focused on user authentication, we investigated three common types of content susceptible to shoulder surfing: text, photos, and PIN authentications. We show that the vulnerability of text and photos depends on the observer’s location relative to the device, while PIN authentications are vulnerable independent of the observation location. We then present PrivacyScout – a novel method that predicts the shoulder-surfing risk based on visual features extracted from the observer’s face as captured by the front-facing camera. Finally, evaluations from our data collection study demonstrate our method’s feasibility to assess the risk of a shoulder surfing attack more accurately.

Despite its importance for assessing the effectiveness of communicating information visually, fine-grained recallability of information visualisations has not been studied quantitatively so far. In this work, we propose a question-answering paradigm to study visualisation recallability and present VisRecall - a novel dataset consisting of 200 visualisations that are annotated with crowd-sourced human (N = 305) recallability scores obtained from 1,000 questions of five question types. Furthermore, we present the first computational method to predict recallability of different visualisation elements, such as the title or specific data values. We report detailed analyses of our method on VisRecall and demonstrate that it outperforms several baselines in overall recallability and FE-, F-, RV-, and U-question recallability. Our work makes fundamental contributions towards a new generation of methods to assist designers in optimising visualisations.

We propose a novel method that leverages human fixations to visually decode the image a person has in mind into a photofit (facial composite). Our method combines three neural networks: An encoder, a scoring network, and a decoder. The encoder extracts image features and predicts a neural activation map for each face looked at by a human observer. A neural scoring network compares the human and neural attention and predicts a relevance score for each extracted image feature. Finally, image features are aggregated into a single feature vector as a linear combination of all features weighted by relevance which a decoder de-codes into the final photofit. We train the neural scoring network on a novel dataset containing gaze data of 19 participants looking at collages of synthetic faces. We show that our method significantly outperforms a mean baseline predictor and report on a human study that shows that we can decode photofits that are visually plausible and close to the observer’s mental image.

We present the first real-world dataset and quantitative evaluation of visual attention of mobile device users in-situ, i.e. while using their devices during everyday routine. Understanding user attention is a core research challenge in mobile HCI but previous approaches relied on usage logs or self-reports that are only proxies and consequently do neither reflect attention completely nor accurately. Our evaluations are based on Everyday Mobile Visual Attention (EMVA) – a new 32-participant dataset containing around 472 hours of video snippets recorded over more than two weeks in real life using the front-facing camera as well as associated usage logs, interaction events, and sensor data. Using an eye contact detection method, we are first to quantify the highly dynamic nature of everyday visual attention across users, mobile applications, and usage contexts. We discuss key insights from our analyses that highlight the potential and inform the design of future mobile attentive user interfaces.

Pursuit eye movements have become widely popular because they enable spontaneous eye-based interaction. However, existing methods to detect smooth pursuits require special-purpose eye trackers. We propose the first method to detect pursuits using a single off-the-shelf RGB camera in unconstrained remote settings. The key novelty of our method is that it combines appearance-based gaze estimation with optical flow in the eye region to jointly analyse eye movement dynamics in a single pipeline. We evaluate the performance and robustness of our method for different numbers of targets and trajectories in a 13-participant user study. We show that our method not only outperforms the current state of the art but also achieves competitive performance to a consumer eye tracker for a small number of targets. As such, our work points towards a new family of methods for pursuit interaction directly applicable to an ever-increasing number of devices readily equipped with cameras.

Common calibration techniques for head-mounted eye trackers rely on markers or an additional person to assist with the procedure. This is a tedious process and may even hinder some practical applications. We propose a novel calibration technique which simplifies the initial calibration step for mobile scenarios. To collect the calibration samples, users only have to point with a finger to various locations in the scene. Our vision-based algorithm detects the users' hand and fingertips which indicate the users' point of interest. This eliminates the need for additional assistance or specialized markers. Our approach achieves comparable accuracy to similar marker-based calibration techniques and is the preferred method by users from our study. The implementation is openly available as a plugin for the open-source Pupil eye tracking platform.

Nowadays, humans are surrounded by many complex computer systems. When people interact among each other, they use multiple modalities including voice, body posture, hand gestures, facial expressions, or eye gaze. Currently, computers can only understand a small subset of these modalities, but such cues can be captured by an increasing number of wearable devices. This research aims to improve traditional human-human and human-machine interaction by augmenting humans with wearable technology and developing novel user interfaces.

This paper proposes a method for achieving accurate ego-vehicle global localization with respect to an approaching intersection; the method is based on the data alignment of the information from two input systems: a Sensorial Perception system, on-board of the ego-vehicle, and an a priori digital map. For this purpose an Extended Digital Map is proposed that contains the detailed information about the intersection infrastructure: detailed landmarks accurately measured and positioned on the map. The data alignment mechanism is thus based on superimposing the sensorial detected landmarks with the corresponding, correctly positioned map landmarks stored in the new Extended Digital Map. The data Alignment Algorithm requires as input, beside the information from the two input systems, the ego-vehicle driving lane. This information is inferred by using a probabilistic approach in the form of a Bayesian Network; the uncertain and noisy character of the sensorial data require such a probabilistic approach in the quest of the ego-lane.