Search

This paper reviews our previous trials of Nudge-XAI, an approach that introduces automatic biases into explanations from explainable AIs (XAIs) with the aim of leading users to better decisions, and it discusses the benefits and challenges. Nudge-XAI uses a user model that predicts the influence of providing an explanation or emphasizing it and attempts to guide users toward AI-suggested decisions without coercion. The nudge design is expected to enhance the autonomy of users, reduce the risk associated with an AI making decisions without users' full agreement, and enable users to avoid AI failures. To discuss the potential of Nudge-XAI, this paper reports a post-hoc investigation of previous experimental results using cluster analysis. The results demonstrate the diversity of user behavior in response to Nudge-XAI, which supports our aim of enhancing user autonomy. However, it also highlights the challenge of users who distrust AI and falsely make decisions contrary to AI suggestions, suggesting the need for personalized adjustment of the strength of nudges to make this approach work more generally., Comment: Accepted at the 9th issue of the International Conference Series on Robot Ethics and Standards (ICRES 2024)

This study considers the attack on reinforcement learning agents where the adversary aims to control the victim's behavior as specified by the adversary by adding adversarial modifications to the victim's state observation. While some attack methods reported success in manipulating the victim agent's behavior, these methods often rely on environment-specific heuristics. In addition, all existing attack methods require white-box access to the victim's policy. In this study, we propose a novel method for manipulating the victim agent in the black-box (i.e., the adversary is allowed to observe the victim's state and action only) and no-box (i.e., the adversary is allowed to observe the victim's state only) setting without requiring environment-specific heuristics. Our attack method is formulated as a bi-level optimization problem that is reduced to a distribution matching problem and can be solved by an existing imitation learning algorithm in the black-box and no-box settings. Empirical evaluations on several reinforcement learning benchmarks show that our proposed method has superior attack performance to baselines.

Transfer learning enhances prediction accuracy on a target distribution by leveraging data from a source distribution, demonstrating significant benefits in various applications. This paper introduces a novel dissimilarity measure that utilizes vicinity information, i.e., the local structure of data points, to analyze the excess error in classification under covariate shift, a transfer learning setting where marginal feature distributions differ but conditional label distributions remain the same. We characterize the excess error using the proposed measure and demonstrate faster or competitive convergence rates compared to previous techniques. Notably, our approach is effective in situations where the non-absolute continuousness assumption, which often appears in real-world applications, holds. Our theoretical analysis bridges the gap between current theoretical findings and empirical observations in transfer learning, particularly in scenarios with significant differences between source and target distributions.

Deep learning models are susceptible to adversarial attacks, where slight perturbations to input data lead to misclassification. Adversarial attacks become increasingly effective with access to information about the targeted classifier. In the context of multi-task learning, where a single model learns multiple tasks simultaneously, attackers may aim to exploit vulnerabilities in specific tasks with limited information. This paper investigates the feasibility of attacking hidden tasks within multi-task classifiers, where model access regarding the hidden target task and labeled data for the hidden target task are not available, but model access regarding the non-target tasks is available. We propose a novel adversarial attack method that leverages knowledge from non-target tasks and the shared backbone network of the multi-task model to force the model to forget knowledge related to the target task. Experimental results on CelebA and DeepFashion datasets demonstrate the effectiveness of our method in degrading the accuracy of hidden tasks while preserving the performance of visible tasks, contributing to the understanding of adversarial vulnerabilities in multi-task classifiers., Comment: 14 pages, 6 figures

Prior work on multilingual sentence embedding has demonstrated that the efficient use of natural language inference (NLI) data to build high-performance models can outperform conventional methods. However, the potential benefits from the recent ``exponential'' growth of language models with billions of parameters have not yet been fully explored. In this paper, we introduce Multilingual Sentence T5 (m-ST5), as a larger model of NLI-based multilingual sentence embedding, by extending Sentence T5, an existing monolingual model. By employing the low-rank adaptation (LoRA) technique, we have achieved a successful scaling of the model's size to 5.7 billion parameters. We conducted experiments to evaluate the performance of sentence embedding and verified that the method outperforms the NLI-based prior approach. Furthermore, we also have confirmed a positive correlation between the size of the model and its performance. It was particularly noteworthy that languages with fewer resources or those with less linguistic similarity to English benefited more from the parameter increase. Our model is available at https://huggingface.co/pkshatech/m-ST5., Comment: Accepted in LREC-COLING 2024

Communication robots have the potential to contribute to effective human-XAI interaction as an interface that goes beyond textual or graphical explanations. One of their strengths is that they can use physical and vocal expressions to add detailed nuances to explanations. However, it is not clear how a robot can apply such expressions, or in particular, how we can develop a strategy to adaptively use such expressions depending on the task and user in dynamic interactions. To address this question, this paper proposes DynEmph, a method for a communication robot to decide where to emphasize XAI-generated explanations with physical expressions. It predicts the effect of emphasizing certain points on a user and aims to minimize the expected difference between predicted user decisions and AI-suggested ones. DynEmph features a strategy for deciding where to emphasize in a data-driven manner, relieving engineers from the need to manually design a strategy. We further conducted experiments to investigate how emphasis selection strategies affect the performance of user decisions. The results suggest that, while a naive strategy (emphasizing explanations for an AI's most probable class) does not necessarily work better, DynEmph effectively guides users to better decisions under the condition that the performance of the AI suggestion is high.

This paper addresses the challenge of selecting explanations for XAI (Explainable AI)-based Intelligent Decision Support Systems (IDSSs). IDSSs have shown promise in improving user decisions through XAI-generated explanations along with AI predictions, and the development of XAI made it possible to generate a variety of such explanations. However, how IDSSs should select explanations to enhance user decision-making remains an open question. This paper proposes X-Selector, a method for selectively presenting XAI explanations. It enables IDSSs to strategically guide users to an AI-suggested decision by predicting the impact of different combinations of explanations on a user's decision and selecting the combination that is expected to minimize the discrepancy between an AI suggestion and a user decision. We compared the efficacy of X-Selector with two naive strategies (all possible explanations and explanations only for the most likely prediction) and two baselines (no explanation and no AI support). The results suggest the potential of X-Selector to guide users to AI-suggested decisions and improve task performance under the condition of a high AI accuracy., Comment: Accepted at the 2024 33rd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN)

A concept-based classifier can explain the decision process of a deep learning model by human-understandable concepts in image classification problems. However, sometimes concept-based explanations may cause false positives, which misregards unrelated concepts as important for the prediction task. Our goal is to find the statistically significant concept for classification to prevent misinterpretation. In this study, we propose a method using a deep learning model to learn the image concept and then using the Knockoff samples to select the important concepts for prediction by controlling the False Discovery Rate (FDR) under a certain value. We evaluate the proposed method in our synthetic and real data experiments. Also, it shows that our method can control the FDR properly while selecting highly interpretable concepts to improve the trustworthiness of the model., Comment: Accepted to IJCAI'23

In this study, we consider a continuous min--max optimization problem $\min_{x \in \mathbb{X} \max_{y \in \mathbb{Y}}}f(x,y)$ whose objective function is a black-box. We propose a novel approach to minimize the worst-case objective function $F(x) = \max_{y} f(x,y)$ directly using a covariance matrix adaptation evolution strategy (CMA-ES) in which the rankings of solution candidates are approximated by our proposed worst-case ranking approximation (WRA) mechanism. We develop two variants of WRA combined with CMA-ES and approximate gradient ascent as numerical solvers for the inner maximization problem. Numerical experiments show that our proposed approach outperforms several existing approaches when the objective function is a smooth strongly convex--concave function and the interaction between $x$ and $y$ is strong. We investigate the advantages of the proposed approach for problems where the objective function is not limited to smooth strongly convex--concave functions. The effectiveness of the proposed approach is demonstrated in the robust berthing control problem with uncertainty.ngly convex--concave functions. The effectiveness of the proposed approach is demonstrated in the robust berthing control problem with uncertainty.

Recent high-$z$ quasar observations strongly indicate that super-Eddington accretion is a crucial phase to describe the existence of supermassive black holes (SMBHs) with $M_\mathrm{BH} \gtrsim 10^9 M_\odot$ at $z \gtrsim 7$. Motivated by the theoretical suggestion that the super-Eddington phase efficiently produces outflows and jets bright in radio bands, we search and find a super-Eddington radio-loud dust-obscured galaxy (DOG) J1406+0102 at $z=0.236$, through cross-matching of the infrared-bright DOGs of Noboriguchi et al. (2019) with the VLA/FIRST 1.4 GHz radio and the SDSS optical spectral catalog. DOG J1406+0102 shows broad components in the Balmer lines. Assuming those lines are from the broad line region, it gives BH mass estimation of $\log\ (M_\mathrm{BH}/M_\odot)=7.30 \pm 0.25$, and AGN luminosity of $\log (L_\mathrm{bol,[OIII]}/\mathrm{erg}~\mathrm{s}^{-1}) = 45.91\pm0.38$ estimated from the intrinsic [OIII] luminosity, resulting in super-Eddington accretion of $\lambda_\mathrm{Edd}\simeq 3$. We show that 1) DOG J1406+0102 is operating strong AGN feedback: the [OIII] outflow velocity exceeds the escape velocity of the host galaxy halo and the kinetic efficiency is obtained as $\approx$ 8% that can be sufficient to quench the host galaxy, 2) the expected future growth pathway of DOG J1406+0102 would join an over-massive BH trajectory and 3) radio-loud DOGs can provide a significant contribution to the high-energy ($\gtrsim$ 100 TeV) cosmic neutrino background if we assume DOG J1406+0102 as a representative of radio-loud DOGs., Comment: 10 pages, 5 figures, submitted to ApJL

For effective collaboration between humans and intelligent agents that employ machine learning for decision-making, humans must understand what agents can and cannot do to avoid over/under-reliance. A solution to this problem is adjusting human reliance through communication using reliance calibration cues (RCCs) to help humans assess agents' capabilities. Previous studies typically attempted to calibrate reliance by continuously presenting RCCs, and when an agent should provide RCCs remains an open question. To answer this, we propose Pred-RC, a method for selectively providing RCCs. Pred-RC uses a cognitive reliance model to predict whether a human will assign a task to an agent. By comparing the prediction results for both cases with and without an RCC, Pred-RC evaluates the influence of the RCC on human reliance. We tested Pred-RC in a human-AI collaboration task and found that it can successfully calibrate human reliance with a reduced number of RCCs., Comment: 8 pages

This study used XAI, which shows its purposes and attention as explanations of its process, and investigated how these explanations affect human trust in and use of AI. In this study, we generated heat maps indicating AI attention, conducted Experiment 1 to confirm the validity of the interpretability of the heat maps, and conducted Experiment 2 to investigate the effects of the purpose and heat maps in terms of reliance (depending on AI) and compliance (accepting answers of AI). The results of structural equation modeling (SEM) analyses showed that (1) displaying the purpose of AI positively and negatively influenced trust depending on the types of AI usage, reliance or compliance, and task difficulty, (2) just displaying the heat maps negatively influenced trust in a more difficult task, and (3) the heat maps positively influenced trust according to their interpretability in a more difficult task., Comment: Published in Proceedings of the 45th Annual Meeting of the Cognitive Science Society (CogSci2023)

We investigate policy transfer using image-to-semantics translation to mitigate learning difficulties in vision-based robotics control agents. This problem assumes two environments: a simulator environment with semantics, that is, low-dimensional and essential information, as the state space, and a real-world environment with images as the state space. By learning mapping from images to semantics, we can transfer a policy, pre-trained in the simulator, to the real world, thereby eliminating real-world on-policy agent interactions to learn, which are costly and risky. In addition, using image-to-semantics mapping is advantageous in terms of the computational efficiency to train the policy and the interpretability of the obtained policy over other types of sim-to-real transfer strategies. To tackle the main difficulty in learning image-to-semantics mapping, namely the human annotation cost for producing a training dataset, we propose two techniques: pair augmentation with the transition function in the simulator environment and active learning. We observed a reduction in the annotation cost without a decline in the performance of the transfer, and the proposed approach outperformed the existing approach without annotation., Comment: The 2022 International Joint Conference on Neural Networks (IJCNN2022)

In this study, we consider simulation-based worst-case optimization problems with continuous design variables and a finite scenario set. To reduce the number of simulations required and increase the number of restarts for better local optimum solutions, we propose a new approach referred to as adaptive scenario subset selection (AS3). The proposed approach subsamples a scenario subset as a support to construct the worst-case function in a given neighborhood, and we introduce such a scenario subset. Moreover, we develop a new optimization algorithm by combining AS3 and the covariance matrix adaptation evolution strategy (CMA-ES), denoted AS3-CMA-ES. At each algorithmic iteration, a subset of support scenarios is selected, and CMA-ES attempts to optimize the worst-case objective computed only through a subset of the scenarios. The proposed algorithm reduces the number of simulations required by executing simulations on only a scenario subset, rather than on all scenarios. In numerical experiments, we verified that AS3-CMA-ES is more efficient in terms of the number of simulations than the brute-force approach and a surrogate-assisted approach lq-CMA-ES when the ratio of the number of support scenarios to the total number of scenarios is relatively small. In addition, the usefulness of AS3-CMA-ES was evaluated for well placement optimization for carbon dioxide capture and storage (CCS). In comparison with the brute-force approach and lq-CMA-ES, AS3-CMA-ES was able to find better solutions because of more frequent restarts.

In the field of reinforcement learning, because of the high cost and risk of policy training in the real world, policies are trained in a simulation environment and transferred to the corresponding real-world environment. However, the simulation environment does not perfectly mimic the real-world environment, lead to model misspecification. Multiple studies report significant deterioration of policy performance in a real-world environment. In this study, we focus on scenarios involving a simulation environment with uncertainty parameters and the set of their possible values, called the uncertainty parameter set. The aim is to optimize the worst-case performance on the uncertainty parameter set to guarantee the performance in the corresponding real-world environment. To obtain a policy for the optimization, we propose an off-policy actor-critic approach called the Max-Min Twin Delayed Deep Deterministic Policy Gradient algorithm (M2TD3), which solves a max-min optimization problem using a simultaneous gradient ascent descent approach. Experiments in multi-joint dynamics with contact (MuJoCo) environments show that the proposed method exhibited a worst-case performance superior to several baseline approaches., Comment: Neural Information Processing Systems 2022 (NeurIPS '22)

Quantum-enhanced (i.e., higher performance by quantum effects than any classical methods) mean value estimation of observables is a fundamental task in various quantum technologies; in particular, it is an essential subroutine in quantum computing algorithms. Notably, the quantum estimation theory identifies the ultimate precision of such an estimator, which is referred to as the quantum Cram\'{e}r-Rao (QCR) lower bound or equivalently the inverse of the quantum Fisher information. Because the estimation precision directly determines the performance of those quantum technological systems, it is highly demanded to develop a generic and practically implementable estimation method that achieves the QCR bound. Under imperfect conditions, however, such an ultimate and implementable estimator for quantum mean values has not been developed. In this paper, we propose a quantum-enhanced mean value estimation method in a depolarizing noisy environment that asymptotically achieves the QCR bound in the limit of a large number of qubits. To approach the QCR bound in a practical setting, the method adaptively optimizes the amplitude amplification and a specific measurement that can be implemented without any knowledge of state preparation. We provide a rigorous analysis for the statistical properties of the proposed adaptive estimator such as consistency and asymptotic normality. Furthermore, several numerical simulations are provided to demonstrate the effectiveness of the method, particularly showing that the estimator needs only a modest number of measurements to almost saturate the QCR bound., Comment: 35 pages, 9 figures. Final version

Evolution strategy (ES) is one of promising classes of algorithms for black-box continuous optimization. Despite its broad successes in applications, theoretical analysis on the speed of its convergence is limited on convex quadratic functions and their monotonic transformation. In this study, an upper bound and a lower bound of the rate of linear convergence of the (1+1)-ES on locally $L$-strongly convex functions with $U$-Lipschitz continuous gradient are derived as $\exp\left(-\Omega_{d\to\infty}\left(\frac{L}{d\cdot U}\right)\right)$ and $\exp\left(-\frac1d\right)$, respectively. Notably, any prior knowledge on the mathematical properties of the objective function such as Lipschitz constant is not given to the algorithm, whereas the existing analyses of derivative-free optimization algorithms require them., Comment: 15 pages

In real-world applications of multi-class classification models, misclassification in an important class (e.g., stop sign) can be significantly more harmful than in other classes (e.g., speed limit). In this paper, we propose a loss function that can improve the recall of an important class while maintaining the same level of accuracy as the case using cross-entropy loss. For our purpose, we need to make the separation of the important class better than the other classes. However, existing methods that give a class-sensitive penalty for cross-entropy loss do not improve the separation. On the other hand, the method that gives a margin to the angle between the feature vectors and the weight vectors of the last fully connected layer corresponding to each feature can improve the separation. Therefore, we propose a loss function that can improve the separation of the important class by setting the margin only for the important class, called Class-sensitive Additive Angular Margin Loss (CAMRI Loss). CAMRI loss is expected to reduce the variance of angles between features and weights of the important class relative to other classes due to the margin around the important class in the feature space by adding a penalty to the angle. In addition, concentrating the penalty only on the important classes hardly sacrifices the separation of the other classes. Experiments on CIFAR-10, GTSRB, and AwA2 showed that the proposed method could improve up to 9% recall improvement on cross-entropy loss without sacrificing accuracy., Comment: 2022 International Joint Conference on Neural Networks (IJCNN 2022)

H1821+643 is the most X-ray luminous non-beamed AGN of $L_\mathrm{14-150 keV}= 5.2\times 10^{45}$ erg s$^{-1}$ in the Swift/BAT ultra-hard X-ray survey and it is also a hyper-luminous infrared (IR) galaxy $L_\mathrm{IR} = 10^{13.2} L_\odot$ residing in the center of a massive galaxy cluster, which is a unique environment achieving the rapid mass assembly of black holes (BH) and host galaxies in the local universe. We decompose the X-ray to IR spectral energy distribution (SED) into the AGN and starburst component using the SED fitting tool CIGALE-2022.0 and show that H1821+643 consumes a large amount of cold gas ($\dot{M}_\mathrm{con}$) with star-formation rate of $\log ( \mathrm{SFR}/M_{\odot}~\mathrm{yr}^{-1}) = 3.01 \pm 0.04$ and BH accretion rate of $\log (\dot{M}_\mathrm{BH}/M_{\odot}~\mathrm{yr}^{-1}) = 1.20 \pm 0.05$. This high $\dot{M}_\mathrm{con}$ is larger than the cooling rate ($\dot{M}_\mathrm{cool}$) of the intra-cluster medium (ICM), $\dot{M}_\mathrm{con}/\dot{M}_\mathrm{cool} \gtrsim 1$, which is one to two order magnitude higher than the typical value of other systems, indicating that H1821 provides the unique and extreme environment of rapid gas consumption. We also show that H1821+643 has an efficient cooling path achieving from $10^7$ K to $10^2$ K thanks to [OIII] 63 $\mu \mathrm{m}$, which is a main coolant in low temperature range ($10^4$ K to $10^2$ K) with a cooling rate of $\dot{M}_{\mathrm{cool}}=3.2\times 10^5\ M_{\odot}\mathrm{~yr^{-1}}$, and the star-forming region extends over 40 kpc scale., Comment: 20 pages, 8 figures, accepted for publication in ApJ

We explore the minimax optimal error associated with a demographic parity-constrained regression problem within the context of a linear model. Our proposed model encompasses a broader range of discriminatory bias sources compared to the model presented by Chzhen and Schreuder (2022). Our analysis reveals that the minimax optimal error for the demographic parity-constrained regression problem under our model is characterized by $\Theta(\frac{dM}{n})$, where $n$ denotes the sample size, $d$ represents the dimensionality, and $M$ signifies the number of demographic groups arising from sensitive attributes. Moreover, we demonstrate that the minimax error increases in conjunction with a larger bias present in the model., Comment: 44 pages, 1figure

We propose CASPER (ChAt, Shift and PERform), a novel dialog system consisting of three types of dialog models: chatter, shifter, and performer. Shifter, which is designed for topic switching, enables a seamless flow of dialog from open-domain chat- to task-oriented dialog. In a user study, CASPER gave a better impression in terms of naturalness of response, lack of forced topic switching, and satisfaction compared with a baseline dialog system trained in an end-to-end manner. In an ablation study, we found that naturalness of response, dialog satisfaction, and task-elicitation rate improved compared with when shifter was removed from CASPER, indicating that topic shift with shifter supports the introduction of natural task-oriented dialog., Comment: 7 pages, 4 figures

Open cloze questions have been attracting attention for both measuring the ability and facilitating the learning of L2 English learners. In spite of its benefits, the open cloze test has been introduced only sporadically on the educational front, largely because it is burdensome for teachers to manually create the questions. Unlike the more commonly used multiple choice questions (MCQ), open cloze questions are in free form and thus teachers have to ensure that only a ground truth answer and no additional words will be accepted in the blank. To help ease this burden, we developed CLOZER, an automatic open cloze question generator. In this work, we evaluate CLOZER through quantitative experiments on 1,600 answers and show statistically that it can successfully generate open cloze questions that only accept the ground truth answer. A comparative experiment with human-generated questions also reveals that CLOZER can generate OCQs better than the average non-native English teacher. Additionally, we conduct a field study at a local high school to clarify the benefits and hurdles when introducing CLOZER. The results demonstrate that while students found the application useful for their language learning. Finally, on the basis of our findings, we proposed several design improvements., Comment: 14 pages, 8 figures

In this study, we investigate the problem of min-max continuous optimization in a black-box setting $\min_{x} \max_{y}f(x,y)$. A popular approach updates $x$ and $y$ simultaneously or alternatingly. However, two major limitations have been reported in existing approaches. (I) As the influence of the interaction term between $x$ and $y$ (e.g., $x^\mathrm{T} B y$) on the Lipschitz smooth and strongly convex-concave function $f$ increases, the approaches converge to an optimal solution at a slower rate. (II) The approaches fail to converge if $f$ is not Lipschitz smooth and strongly convex-concave around the optimal solution. To address these difficulties, we propose minimizing the worst-case objective function $F(x)=\max_{y}f(x,y)$ directly using the covariance matrix adaptation evolution strategy, in which the rankings of solution candidates are approximated by our proposed worst-case ranking approximation (WRA) mechanism. Compared with existing approaches, numerical experiments show two important findings regarding our proposed method. (1) The proposed approach is efficient in terms of $f$-calls on a Lipschitz smooth and strongly convex-concave function with a large interaction term. (2) The proposed approach can converge on functions that are not Lipschitz smooth and strongly convex-concave around the optimal solution, whereas existing approaches fail., Comment: accepted for GECCO 2022