In human-AI decision making, designing AI that complements human expertise has been a natural strategy to enhance human-AI collaboration, yet it often comes at the cost of decreased AI performance in areas of human strengths. This can inadvertently erode human trust and cause them to ignore AI advice precisely when it is most needed. Conversely, an aligned AI fosters trust yet risks reinforcing suboptimal human behavior and lowering human-AI team performance. In this paper, we start by identifying this fundamental tension between performance-boosting (i.e., complementarity) and trust-building (i.e., alignment) as an inherent limitation of the traditional approach for training a single AI model to assist human decision making. To overcome this, we introduce a novel, human-centered adaptive AI ensemble that strategically toggles between two specialist AI models—the aligned model and the complementary model—based on contextual cues, using an elegantly simple yet provably near-optimal Rational Routing Shortcut mechanism. Comprehensive theoretical analyses elucidate why the adaptive AI ensemble is effective and when it yields maximum benefits. Moreover, experiments on both simulated and real-world data show that when humans are assisted by the adaptive AI ensemble in decision making, they can achieve significantly higher performance than when they are assisted by single AI models that are trained to either optimize for its independent performance or even the human-AI team performance.

In Deep Neural Networks (DNNs), manipulating gradients is central to various algorithms, including data subset selection and instance attribution. For better tractability, practitioners often resort to using only the gradients of the last layer as a heuristic, instead of the full gradient across all model parameters, which we show is detrimental due to the Support Vector Effect (SVE). We introduce SVE, a max-margin-like behavior in the last layer(s) of DNNs and employ it to thoroughly scrutinize prevalent data selection and attribution methods relying on last layer gradients. Our investigation exposes limitations in these techniques and not only provides explanations for previously observed pitfalls, like lack of diversity and temporal performance degradation, but also offers fresh insights, including the vulnerability of existing methods to basic poisoning attacks and the potential for competitive performance using much simpler alternatives. Based on insights from SVE, we craft new methods RandE and PAE for data subset selection and instance attribution, respectively, which often outperform the purported state-of-the-art at a fraction of the cost, emphasizing the practical advantages of more efficient and less complex approaches.

Phishing attacks remain pervasive and continue to be a source of significant monetary loss, identity theft, and malware. One of the challenges is that in most organizational settings, the detection paradigm is inherently about identifying and reacting to threats in real-time, as they are unfolding. As a way to complement these efforts with greater foresight, we introduce the idea of phishcasting — forecasting of phishing threat levels weeks or months into the future. Given that phishing attack volume time series data is noisy and devoid of traditional seasonal and cyclical trends, we extend the time series forecasting framework to utilize multiple time series, auxiliary information and alternate representations. We also introduce CoT-Net, a flexible, end-to-end CNN-LSTM based deep learning method for forecasting of complex phishing attack volume time series. CoT-Net uses time series embeddings to uncover correlations between organizational attack patterns within and across industry sectors. Using a publicly available test bed featuring multiple organizations’ attack volume over time, we find CoT-Net to outperform most state-of-the-art time series forecasting methods. By showing that phishcasting might be possible and practical, our work has important proactive implications for cybersecurity.