AI assisted job applications (and application screening)

 NBC news has the story:

AI is supposed to make applying to jobs easier — but it might be creating another problem
Companies have touted new AI technology that allows users to apply to thousands of jobs per day, flooding job openings with resumes.  By Sophia Pargas

"With the help of new AI software, applicants can now revise their resumes and cover letters, receive live interview scripts and launch chatbots to submit thousands of applications almost instantly.

...

"LazyApply, Simplify and AI Hawk are all AI services that launch assistants to help collect applicant information and submit automated job applications through sites like LinkedIn and Indeed. Using the tools, job seekers can save hundreds of hours — and apply to up to thousands of jobs a day, according to the services.

"AI Apply, which claims its users are “80% more likely to get hired,” offers tools like a cover letter and resume builder, an auto-apply feature, an interview practice generator and a specialized interview buddy. A premium membership starts at $38 a month, according to the website — a cost AI Apply notes is much lower and more accessible than the cost of a career counselor.

...

"According to the company’s respective websites, OfferGoose has helped applicants land over 17,000 job offers, and AI Apply has over 340,000 “freelancers and job seekers” using the service.

"Nevertheless, many companies have enlisted additional safeguards to detect the use of AI. Some employers now require e-confirmation codes to submit applications and prevent automated submissions. Others have added prompts asking AI bots to use specific words like “banana” in responses to catch chat-generated application answers, according to longtime job recruiter and tech agency director of talent Mike Peditto.

"But job seekers are not the only party using AI in the application process, according to a recent University of Washington study. The researchers estimated that “99% of Fortune 500 companies are already using some sort of AI assistance when making hiring decisions,” and found that resume-screening tools have an incredible bias toward white, male applicants — with the large language models favoring white-associated males roughly 85% of the time and disadvantaging Black males in “up to 100% of cases.”

"The use of AI by both applicants and employers has created a distinct dynamic, in which applicants are using AI, but are also reticent about employers screening them with AI."

What preferences are revealed by market designs? by Oğuzhan Çelebi (who is on the job market)

Oğuzhan Çelebi is on the job market this year (listed on the job market websites of both MIT and Stanford where he's finishing a two year postdoc). He has a new paper that takes a really novel approach to market design.  It looks at the implicit preference relation (if one exists) that a mechanism reveals when a revealed preference analysis is done of the choices it makes from different possible menus of alternatives. The specific focus is on preferences for diversity revealed by mechanisms used in connection with affirmative action.

Diversity Preferences and Affirmative Action, by Oğuzhan Çelebi 

Abstract: "In various contexts, institutions allocate resources using rules that determine selections given the set of candidates. Many of these rules feature affirmative action, accounting for both identity and (match) quality of individuals. This paper studies the relationship between these rules and the preferences underlying them. I map the standard setting of market design to the revealed preference framework, interpreting choice rules as observed choices made across different situations. I provide a condition that characterizes when a rule can be rationalized by preferences based on identities and qualities. I apply tests based on this condition to evaluate real-world mechanisms, including India’s main affirmative action policy for allocating government jobs, and find that it cannot be rationalized. When identities are multidimensional, I show that non-intersectional views of diversity can be exploited by dominant groups to increase their representation and cause the choice rules to violate the substitutes condition, a key requirement for the use of stable matching mechanisms. I also characterize rules that can be rationalized by preferences separable in diversity and quality, demonstrating that they lead to a unique selection within the broader set of policies that reserve places based on individuals’ identities."

########

He's a target of opportunity for any department interested in market design.

Job market for Economics Ph.D.s: 2024 versus the past 5 years

 Here's a scary story for Halloween: the job market for new Econ PhDs is off to a somewhat slow start: in the figures below, see the bright blue line of positions for 2024 to date.  There's still room to grow (as indeed we've seen in previous years).  And we're way ahead of 2020, the deep Covid year.

To: Members of the American Economic Association
From: AEA Committee on the Job Market:  John Cawley (chair), Elisabeth “Bitsy” Perlman, Al Roth, Peter Rousseau, Wendy Stock, and Stephen Wu
Date: October 30, 2024
Re: JOE job openings by sector, 2024 versus the past 5 years

"This memo reports the cumulative number of unique job openings on Job Openings for Economists (JOE), by sector and week, compared to the same week in recent years."