Theory
Through natural language processing and LLM we can use a process called forensic semiotics to construct sign systems by which we can deconstruct and interpret candidate resumes and fit them to the model requirements defined by the hiring manager.
In theory, a product like “Microsoft Project” might appear in many sign forms in a resume. It could be just “Project” or it could appear as “MS Project” or “Microsoft Project”. “Project” is the toughest one since it has the least context and could appear elsewhere in the resume for which the context has limited value. (In these cases, recruiter heuristics suggest creating a sign which combines “Project” with another likely piece of Project management software from the same suite (Visio) to create a sign: (project AND Visio) which serves to reduce false positives and increase the probable contextuality of the term in the resume.)
Or it could be just generally some kind of PPM and then we can dump a bunch of terms in a boolean query like (“MS Project” OR “Microsoft Project” OR (Project AND Visio) OR Clario OR Planview) etc. This entire sequence in quotes becomes our “sign” for PPM.
Applying a view which conceives of a series of boolean “signs” as a sequence in a search string, we can approach a database of resumes to identify those resumes which are a match for the PPM sign system and ignore the resumes which do not contain the signs.
The semiotic boolean approach has advantages over a semantic approach because the results are precise.
If the recruiter is inexperienced, they may benefit from a semantic search or semantic search suggestions; but in general, an experienced recruiter armed with Boolean as a sign system and a knowledge of the organizational culture and the job requirements and hiring manager will be much more equipped to get exactly what they are looking for.
Boolean criteria as signs become objective criteria from which our data set of resumes can be abstracted from the set of all resumes
Once we have an abstracted set of precise resumes that accurately reflect the boolean sign requirements, we can perform more subjective analysis on those resumes to weigh them against one another and determine fit with the overall model on a kind of % match basis
This approach can be refined for the unique job and requirements (for example a job where certification may be weighed higher than education)
Principles
Blindness to background focus on fit of experience to the requirement as we fit model to data
Belief that by focusing on blind principles in hiring we will eventually build a employee population which is a model of the real world society at scale
Responsiveness, kindness, and engagement are essential for keeping the best candidates “hot” and ready to accept the job if offered
Reciprocally, despite most people saying they want to get feedback if not selected for closure, it is not always best to give people negative feedback unless they specifically ask to know their disposition. In these cases, it is ethical to tell them that they were not selected quickly, while apologizing for not letting them know sooner. Otherwise, never inflicting this “psychic wound” on good candidates makes it easier to work with them again in the future. They can abduct they were not hired.
The recruiter can tell them if there is feedback that is positive, they will let them know right away; but also that they will not be contacting them if they do not have news. This can imply to the candidate the recruiter likes them but they may not hear from the recruiter if they were not selected. However, if the person does reach out and really wants the psychic wound inflicted which they should already have abducted; then the recruiter should respond swiftly and kindly in explaining why. Never go down a rabbit hole with the candidate especially when they are upset by the outcome. It’s just the way it is and it isn’t the recruiter decision. General comments about improving interview performance which were of concern in the interview, such as a focus on practicing STAR responses or improving brevity in responses can improve the candidate model without divulging specific areas of feedback which may be harmful to the candidate model’s cognitive process and result in negative feelings directed at the recruiter or (former employer).
Apply Cialdini’s concepts of Influence in an ethical framework which is not obvious and is not sappy or aggressive. Be open to discussing Cialdini openly if asked what principles underlie the psychological factors behind the model as well as cognitive forensic computational semiotics.
Ignore most non requirement aspect of the job and look at boolean keyword/keyphrase fit to model requirements
Ensure framework is in place so that all actual requirements are explicitly in the job description and not implicit on the part of the hiring manager
Model can abduct missing requirements by determining gap between past hire and job description
Scoring
(For experienced candidates) Weigh company in a model but at a relatively moderate percentage of the evaluation
(For all candidates) Weigh education in the model but at a relatively moderate percentage with a focus on the objective requirements for the degree and role requirements rather than the institution from which the degree was received
(For experienced candidates) Weigh the average job duration as a significant portion of the model (eg. greater than moderate weight but not maximum weight), and especially as it relates to career progression, working successfully in various organizations, taking on new roles; and demonstrating consistent interaction with the core skills and competencies required for the role.
(For all candidates) Weigh certification(s) in the model, but at a relatively low weight unless it is a specific requirement for the role in which case weigh it at high/maximum value.
** Over certification may indicate careerist focus when accompanied by average short jobs
(For all candidates) Look for responsiveness with the recruiter as a key indicator of their engagement with the model
Updates:
Consider periodic “cold close” as part of the model to highly ranked candidates in order to assess candidate engagement and likelihood of future offer acceptance
** Prompt to user:”Looks like you had a good interview. Following the interview can you see yourself in the role?”
* Consider other integration with Recruiter toolkit for referrals, etc.
Prototype Approach:
Consider an approach to refining boolean queries based on a list of keywords in database and resume set abstraction
Train the model to construct efficient boolean queries which model the recruiter input
Input the queries into existing tools
Test on Applicant Tracking Database Resume set
An individual skill within the PPM set could be weighed higher than other technologies based on manager preferences (eg exact match vs product analog)
(note: I wrote this before I lost my recruiting job last year and shared with my employer. Just figured I’d throw it out there.)