通过与深度神经网络的解释交互，使其以正确的科学理由获得正确结果
Making deep neural networks right for the right scientific reasons by interacting with their explanations
AI总结：
针对深度神经网络易依赖数据集混淆因素的“聪明汉斯”问题，提出解释性交互式学习XIL范式，让科学家通过反馈模型解释参与训练修正，可避免该问题并合理调控对模型的信任。
摘要：
深度神经网络在诸多实际应用中已展现出优异性能。遗憾的是，它们可能表现出类似“聪明汉斯（Clever Hans）”的行为——利用数据集中的混淆因素来取得高性能。本研究提出了全新的“解释性交互式学习（explanatory interactive learning，XIL）”学习范式，并在植物表型研究任务中验证了其优势。XIL将科学家纳入训练循环，由科学家针对模型的解释提供反馈，从而交互式地修正原始模型。实验结果表明，XIL有助于避免机器学习中的“聪明汉斯”现象，并能增强（或在适当时削弱）对底层模型的信任。
英文摘要：
Deep neural networks have shown excellent performances in many real-world applications. Unfortunately, they may show "Clever Hans"-like behavior -- making use of confounding factors within datasets -- to achieve high performance. In this work, we introduce the novel learning setting of "explanatory interactive learning" (XIL) and illustrate its benefits on a plant phenotyping research task. XIL adds the scientist into the training loop such that she interactively revises the original model via providing feedback on its explanations. Our experimental results demonstrate that XIL can help avoiding Clever Hans moments in machine learning and encourages (or discourages, if appropriate) trust into the underlying model.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
