Is there good reason to think that optimizing resumes will help with the job searching process?
I've heard that many resume screens are automated these days, which has led to my hearing stories of people trying to game this by cramming keywords in invisible font and other things like that.
Can anyone with experience in recruiting shed some light on this matter?
Agreed.
After writing my own static site generator in Python, I went back and was impressed with all the options available in Pelican. Lots of things I wished I had built were included.
Doesn't seem like a well-curated list.
Some of these, e.g. Uber's God Mode, don't even fall under the normal usage of the term "AI".
Others, like Palantir or WeChat, seem to denigrate an entire company (rather than specific practices).
And the racist chatbot thing isn't even an actual product, nor was it intentionally designed to be racist.
Dustin is also notable for starting Good Ventures (https://www.goodventures.org/about-us) which funds a lot of research and philanthropy into some great causes.
I see. Thanks for adding to the clarification. I think that the presentation of nulls as "results" can definitely be disingenuous. Ideally, science would have a better database to keep track of what people find, where we could add nulls in a way that doesn't highlight their "importance". As the person above says, reporting nulls is still useful to prevent p-hacking and publication bias.
(Of course, ideally I think we'd be better off focusing on reporting the data in a Bayesian approach, but that hasn't really gotten traction in the broader community.)
I definitely think there's more room for this sort of guided / ML analysis, but I'm not quite sure to make traction on extracting the structure of scientific papers...hopefully someone with more experience can chime in.
Fogg has some good thoughts on positive behavior change. James Clear, who wrote Tiny Habits, is another good person in this space who writes reasonable stuff.
Self-plug for an overview on research into the habit formation literature that I recently cleaned up. Covers many evidence-backed interventions (of varying quality) for habit formation and removal.
Overall, I think parts of this approach might throw the baby out with the bathwater. Even if we do restructure the way that society works, presumably we'll still care about interventions to improve well-being. I think effective altruism (https://www.effectivealtruism.com/) has provided a lot of traction on how to think about these issues. Even if you disagree with their moral stance, I think it's hard to deny that people in the EA community have been very thoughtful about quantifying interventions and thinking about impact.
And I don't think these sorts of questions are going to go away, come a social restructuring. There seems to be a lot that we can salvage from the EA viewpoint, even if we decide systemic change is the way forward, and I think that's worth emphasizing.
With the rise of highly effective black box models, there's been a rise in interest towards interpretability for machine learning. This is an area I'm excited about, so I did a dive into the existing research.
I found a few surprising things (lack of user studies and some explanations even "working" on randomly initialized models), and I wanted to share a glimpse into the field for others who might be interested.
There's definitely a lot more room to explore, be it from the usability standpoint or from the more technical standpoint. Hopefully this can be an accessible jumping off point for others trying to enter the conversation.
- "only 39 scientists from 7 countries have been subject to criminal sanctions between 1979 and 2015 (Oransky and Abritis, 2017)" That seems...very low.
- "The Retraction Watch database—the largest of its kind—currently includes more than 18,500 retracted articles (Retraction Watch database, 2019). A recent analysis of 10,500 retracted papers up to 2016 showed that 0.04% of papers are retracted." This is once again a lower-bound; presumably if you account for additional authors and p-hacking the numbers go up a lot.
Pushing for replication and improved methodology can help, but some of these issues seem to be related to scale. There are many more people outputting papers than there are people willing to vet them (outside of peer review). Furthermore, when you have many people researching hot fields, you should expect false positives and overestimates to dominate published results, even when everyone is trying to practice good statistical hygiene. (https://journals.plos.org/plosmedicine/article?id=10.1371/jo...)
I've heard that many resume screens are automated these days, which has led to my hearing stories of people trying to game this by cramming keywords in invisible font and other things like that.
Can anyone with experience in recruiting shed some light on this matter?