AI Has a Better Resume Than You
1st March 2021
AI Has a Better Resume Than You Banner
AI Has a Better Resume Than You Banner
AI Has a Better Resume Than You Banner

For the time being, however, hiring remains a largely human process. We tinker with our resumes, hoping to impress through the careful presentation of our credentials. AI tools may soon transform industries, but they don't have what it takes to appeal to real hiring managers.

Or do they?

We wondered what would happen if we asked AI to prepare impressive resumes, using real examples to train a machine learning model. Could the computer invent convincing and qualified candidates, creating fictional resumes matched to demanding job descriptions?

Focusing on the competitive technology industry, we asked an AI model to produce nine distinct resumes for candidates in a range of developer roles. Our results ranged from odd to impressive and came uncomfortably close to mimicking an attractive candidate's credentials.

To explore the results of our experiment, you can view the AI-generated resumes right here. Below, we'll offer a full description of how we conducted this experiment and discuss the implications of our work for professionals and businesses around the world.

Our Experiment: Generating Resumes

The process for generating these resumes was surprisingly straightforward. We utilised OpenAI's GPT2 pretrained 774M model and fine-tuned (did additional training) with roughly 1,000 technology worker resumes scraped from various online sources. The model was trained to 5,000 steps (a relatively small amount of fine-tune training) and was yet still able to produce impressive results. Formatting and design of the resumes was done by human designers. The faces of the applicants in each resume were generated by StyleGan2, an AI human face generating algorithm.

Learning from the Models: Reversing AI Resumes

Clearly, the resumes produced by our experiment are hardly perfect: Some of the verbiage used is odd or unclear, and many elements appear in improper categories. But in many respects, the resumes resulting from our approach are strikingly close to what a real candidate's CV might look like. As AI technologies rapidly evolve, it's safe to assume that even more convincing facsimiles will soon be possible.

Accordingly, what are the practical implications of our experiment for businesses and professionals around the world? One interesting possibility involves what we can learn from AI-generated resumes, rather than how well we can teach them to imitate us.

Consider your own career trajectory: Do you have a dream job in mind or specific aspirations to obtain a more elevated position? Maybe you're seeking to change careers or enter a new role within your industry. Whatever your own professional hopes, AI could provide highly instructive models.

Take our experiment's examples: These resumes are full of credentials necessary to obtain coveted tech positions, from specific programming languages to prior positions and degrees. These pieces of information can be crucial in guiding your own professional advancement. By understanding the perfect AI-generated candidate, you can identify room for improvement in your own resume – and burnish your own credentials accordingly.

Additionally, machine text generation reflects the specific language used within a given sector. Using the right jargon is one way to indicate your expertise, and AI can show you which terms and phrases recur in relevant resumes. This information could be invaluable in deciding how to frame your experience by emphasising your qualifications using the right terminology.

Artificial Applicants: An Unethical Edge?

In recent years, a roiling global economy has generated demand for qualified workers, leading to significant talent shortages across several industries. These talent shortfalls are especially acute in technology fields, as companies across the world have struggled to find and hire workers with crucial technical skills.

This dynamic has given rise to cutthroat competition: From lean startups to the juggernauts of Silicon Valley, businesses are aggressively recruiting tech professionals – and poaching them if necessary. Even as COVID-19 has upended the global workforce, giants such Apple, Google, and Facebook have sustained rapid recruitment campaigns.

In a world where tech talent is at a premium, it's not hard to imagine that resumes like the ones we created would catch the eye of eager hiring managers. After all, these AI-generated candidates possess a host of enviable skills and credentials, some of which are in particularly high demand. But this reality suggests another possible implication of our experiment. Could this kind of deception be employed for shady purposes, confusing or delaying a rival company's hiring process?

Imagine the incentives for a ruthless corporate actor: If you and your competitors are hiring from the same talent pool, you could use AI technologies to flood their recruitment channels with fake resumes. With your rival sorting through tons of fictitious candidates, you can focus on recruiting the real ones.

These devious tactics would be more than a minor annoyance: HR professionals currently find hiring skilled professionals quite challenging, and a rash of AI-generated imposters might place impossible demands on their limited time and resources.

Moreover, many companies are easing barriers for applicants, streamlining their hiring processes to attract as many candidates as possible. But a flurry of fake resumes might require companies to take additional screening steps, therefore deterring real applicants.

Would companies stoop low enough to target their competitors in this way? The answer depends on the scruples of the businesses involved and the intensity of the war for talent. But these duplicitous tactics needn't become common practice to cast a pall over all hiring. Once the threat of AI-generated resumes is widely understood, companies will have little choice but to consider candidates more skeptically.

HR Goes AI: How Machine Learning Is Transforming Hiring

While our experiment explores a novel application of AI, it's hardly the first time these technologies have intersected with the hiring process. In increasingly varied ways, companies are utilising AI to innovate their search for talent.

These advances are already reshaping how applicants are identified and assessed and may soon determine the fortunes of the workforce. Accordingly, it's essential for applicants to understand how these technologies are being utilised in order to navigate this shifting landscape to their advantage. Here are just a few powerful implementations of AI in the hiring process:

Application Evaluation: AI can be trained to seek out specific terms in cover letters and resumes, passing along candidates who meet pre-established criteria. But companies can also use more sophisticated AI tools to assess applicants, using models trained on data from successful employees. By considering the characteristics of successful current employees, these models can predict a candidate's likelihood of succeeding at the company. Unfortunately, such models can be governed by bias: They're unlikely to identify new sources of talent and may actively discriminate against diverse applicants.

Personality Testing: Personality can't be successfully conveyed in a written application, so hiring managers often rely on interviews to assess desirable characteristics, such as emotional intelligence and strong communication skills. But AI technology can introduce hard data into these calculations, administering personality measures and analysing their results. In many cases, candidates answer a few questions or complete a brief exercise. AI tools then scrutinise their performance, producing data-based predictions about their personalities rather than the "gut feelings" hiring managers have historically relied on.

Candidate Outreach: Companies increasingly utilise AI tools to engage their customers, improving many aspects of their experience as a result. When customers interact with AI "bots," they often receive more immediate and effective service than a human employee could offer. Similarly, many businesses are using AI tools in their hiring processes, eliminating much of the delay and uncertainty that can characterise hiring. While engaging with a bot may seem off-putting to some, these programs can handle key aspects of communication, from relaying updates to scheduling interviews.

Clearly, these methods and our experiment represent just some of the ways in which AI will continue to transform hiring – for companies and candidates alike. Though the full trajectory of these technologies remains to be revealed, we can be certain that their prominence will increase in the years to come. As you consider your own career prospects, will you be ready for the age of AI hiring?

Here's one way to prepare yourself for a competitive hiring market: Become an expert in the technologies that will define the future of business. To do so, you'll need to embrace education and training, investing in your skills to ensure your success. This won't necessarily be easy: It can be tough to balance building your resume with the demands of daily life.

Let us be your ally in advancing your career. At James Cook University, we specialise in programs of study suited to the modern professional, whether you're already working or building a foundation for a future career. Our Master of Data Science provides the support needed to open a wealth of opportunities in a data-driven world, putting you ahead of the data science revolution. We offer flexible online study options to empower you to gain the knowledge and credentials you need, on your terms and your schedule. Explore our programs to learn how we can help you to stand out from the crowd of candidates – AI-generated or otherwise.

Fair Use Statement

We hope our results will intrigue (and unsettle) a broad range of readers, whether or not they work directly with AI technologies. Accordingly, we welcome you to share this content widely so long as you do so for noncommercial purposes. Please also link back to this page whenever you cite our work, allowing others to access and enjoy the full project.

Data Science

Related study options