Multi-hop Question Generation is the task of generating questions which require the reader to reason over and combine information spread across multiple passages using several reasoning steps. Chain-of-thought rationale generation has been shown to improve performance on multi-step reasoning tasks and make model predictions more interpretable. However, few-shot performance gains from including rationales have been largely observed only in +100B language models, and otherwise require large scale manual rationale annotation. In this work, we introduce a new framework for applying chain-of-thought inspired structured rationale generation to multi-hop question generation under a very low supervision regime (8- to 128-shot). We propose to annotate a small number of examples following our proposed multi-step rationale schema, treating each reasoning step as a separate task to be performed by a generative language model. We show that our framework leads to improved control over the difficulty of the generated questions and better performance compared to baselines trained without rationales, both on automatic evaluation metrics and in human evaluation. Importantly, we show that this is achievable with a modest model size.

Social biases based on gender, race, etc. have been shown to pollute machine learning (ML) pipeline predominantly via biased training datasets. Crowdsourcing, a popular cost-effective measure to gather labeled training datasets, is not immune to the inherent social biases of crowd workers. To ensure such social biases aren't passed onto the curated datasets, it's important to know how biased each crowd worker is. In this work, we propose a new method based on counterfactual fairness to quantify the degree of inherent social bias in each crowd worker. This extra information can be leveraged together with individual worker responses to curate a less biased dataset.

Crowdsourcing is a popular methodology to collect manual labels at scale. Such labels are often used to train AI models and, thus, quality control is a key aspect in the process. One of the most popular quality assurance mechanisms in paid micro-task crowdsourcing is based on gold questions: the use of a small set of tasks of which the requester knows the correct answer and, thus, is able to directly assess crowd work quality. In this paper, we show that such mechanism is prone to an attack carried out by a group of colluding crowd workers that is easy to implement and deploy: the inherent size limit of the gold set can be exploited by building an inferential system to detect which parts of the job are more likely to be gold questions. The described attack is robust to various forms of randomisation and programmatic generation of gold questions. We present the architecture of the proposed system, composed of a browser plug-in and an external server used to share information, and briefly introduce its potential evolution to a decentralised implementation. We implement and experimentally validate the gold detection system, using real-world data from a popular crowdsourcing platform.  Our experimental results show that crowdworkers using the proposed system spend more time on signalled gold questions but do not neglect the others thus achieving an increased overall work quality. Finally, we discuss the economic and sociological implications of this kind of attack.