System-wide educational reforms are difficult to implement in the United States, but despite the difficulties, reforms can be successful, particularly when they are associated with broad public support. This study reports on the nature of the public sentiment expressed about a nationwide science education reform effort, the Next Generation Science Standards (NGSS). Through the use of data science techniques to measure the sentiment of posts on Twitter about the NGSS (N = 565,283), we found that public sentiment about the NGSS is positive, with only 11 negative posts for every 100 positive posts. In contrast to findings from past research and public opinion polling on the Common Core State Standards, sentiment about the NGSS has become more positive over time—and was especially positive for teachers. We discuss what this positive sentiment may indicate about the success of the NGSS in light of opposition to the Common Core State Standards.

The growing capability and availability of generative language models has enabled a wide range of new downstream tasks. Academic research has identified, quantified and mitigated biases present in language models but is rarely tailored to downstream tasks where wider impact on individuals and society can be felt. In this work, we leverage one popular generative language model, GPT-3, with the goal of writing unbiased and realistic job advertisements. We first assess the bias and realism of zero-shot generated advertisements and compare them to real-world advertisements. We then evaluate prompt-engineering and fine-tuning as debiasing methods. We find that prompt-engineering with diversity-encouraging prompts gives no significant improvement to bias, nor realism. Conversely, fine-tuning, especially on unbiased real advertisements, can improve realism and reduce bias.

No abstract

As the use of social media increases in daily life, it has also increased for institutions in the field of education. While there may be benefits for schools to use this media outlet, the privacy of students within those schools may be at risk when their names and photos are shared on such a publicly accessible domain. In this study, we analyzed the extent to which students’ privacy is protected by qualitatively coding a random sample of 100 Facebook posts made by U.S. school districts from a population of over 9.3 million photo posts that we collected. Using inferential techniques, we found that students are somewhat protected compared to teachers and community members, with only 2.67% of students’ detected faces able to be identified by name. These numbers at first appear small, but if applied to the entire population, this could potentially leave between 153,218 and 1,l53,844 students identifiable to anyone on the Internet; the number of photos of students posted by schools and districts is much greater still, between 15.2 and 20.3 million. The same measure for staff and community members were 4.6% and 16%, respectively. We discuss the severity and scale of these privacy threats and make recommendations for research on student privacy in social media and other informal education-related contexts. In all, these could represent the largest publicly available collection of identifiable photos of students (and children) in the United States and could seriously threaten the privacy of those identified.

No abstract