AI in Recruiting
D r. D a n i e l M ü h l b a u e r
T H E G O O D , T H E B A D A N D T H E U G L Y
Knowing vs. Doing
currently use or implement AI projects in HR.
think AI is a highly relevant technology.
79% of firms 8% of firms
AI-study 2020 by Deloitte
PEOPLE
TOOLS
DATA
ORG
PROCESS
Robot-Coworker = popular?
of the surveyed people are
optimistic and excited about
robot-coworkers (n=8000)
53% 64% 36%
AI@WorkReport by Oracle
of the surveyed people
would trust a robot more
than their manager (n=8000)
of the surveyed people think
that robots are less biased
than managers (n=8000)
MALE APPLICANTS
RECEIVE SIGNIFICANTLY
MORE RECRUITER
CALLBACKS THAN
FEMALE APPLICANTS.
(Rivera/Tilcsik, 2016)
MALE APPLICANTS
ARE PREFERRED FOR
"MALE
OCCUPATIONS", BUT
WOMEN ARE NOT
PREFERRED FOR
"FEMALE
OCCUPATIONS".
(Koch et al., 2015)
NO REVERSAL OF
DISCRIMINATION IN
RECRUITING AFRICAN-
AMERICAN
APPLICANTS SINCE
1989.
(Quillian et al., 2019)
FEMALE
APPLICANTS OVER
40 RECEIVE
SIGNIFICANTLY
FEWER RECRUITER
CALLBACKS.
(Carlsson et. al. 2019)
APPLICANTS WITH
A MIGRATION
BACKGROUND
RECEIVE
SIGNIFICANTLY
FEWER
RECRUITER
CALLBACKS.
(Moa, 2017)
FEMALE
RECRUITERS
PREFER WOMEN.
MALE
RECRUITERS DO
NOT PREFER MEN.
(Carlsson/Eriksson,
2019)
WOMEN
GENERALLY
RECEIVE FEWER
RECRUITER
CALLBACKS,
ESPECIALLY IF
THEY HAVE
CHILDREN.
(Gonzales et al., 2019)
SHOULD WE
PERHAPS GIVE
INTELLIGENT HR
TECHNOLOGY A
CHANCE AFTER
ALL?
RECRUITING FAILS?
46% of all
hires
The solution to the hiring puzzle
is finding your firm’s happy place
between HR Tech and HR Touch!
DANIEL MÜHLBAUER
THE GOOD
RECRUITING ANALYTICS &
PROCESS MINING
Intelligent technology analyzes your
recruiting funnel and identifies
errors, circularities, inefficiencies of
your recruiting.
INTELLIGENT STRING-
BUILDER
Intelligent technology analyzes your
job posting and suggests the best
Boolean string for active sourcing.
JOB-ADS w/ COMMUTE-
ANALYSIS
For each job posting, candidates
receive an analysis of the work route
including seasonal weather effects,
means of transport and rush-hours.
TEXT-MINING OF
COMPANY INFOS
An intelligent technology analyzes all
publicly available company
information and derives various
indicators of employer quality from
the language used.
Good AI-based REC technology focuses
specific HR tasks and enables a linkage of
artificial and human intelligence!
THE GOOD
SKILL-BASED MATCHING
Intelligent technology compares
requirements and qualifications at
the ski level and predicts a match
quality.
AUTOMATED APTITUDE
TESTING
Intelligent technology combines
validated methods of aptitude
diagnostics to predict aptitude in real
time.
PERSONALIZED
CAREERPAGES
Intelligent technology modifies
career site content based on target
group analysis and typical click
behavior.
AVATARS FOR RECRUITING
TRAINING
Intelligent virtual avatars enable the
realistic simulation of interviews for a
variant-rich training of recruiting trainees.
Good AI-based REC technology focuses
specific HR tasks and enables a linkage of
artificial and human intelligence!
THE BAD
EMOTION RECOGNITION
Emotion recognition is a very serious
research direction, but its results
have not yet reached market
maturity.
FIT-PREDICITION BASED
ON X
The fit between people, jobs, teams,
companies or cultures are very
complex. The data models are often
not reliable.
AD TARGETING ON SOCIAL
MEDIA
Job posting algorithms are not
transparent enough. They can subtly
discriminate.
Poor AI-based REC technology focuses on
contexts that are too complex, are
opaque, or are still in validation. As a
result, their use should not yet be full-
scale, but in pilots or validation studies.
THE UGLY
INVALID
SELECTIONMETHODS
Poor selection processescannot be
salvaged with AI use. Examples:
Graphology, MBTI, DISG, profiling.
DISCRIMINATION IN
VIDEO RECRUITMENT
Algorithms for image and video
processing often work significantly
worse for non-light-skinned people:
PERSONALITY-ANALYSES
-> PERFORMANCE
Personality is not a stable predictor
of subsequent job performance. AI-
based technology does nothing to
change that.
Bad AI-based REC technology is based on
dubious or even discriminatory
procedures whose automation reinforces
systematic disadvantage!
FACE-ANALYSES -> ATTITUDES
& BEHAVIOR
Analyses of micro-expressionsto
infer any traits or emotions of a
person are not valid.
YOUR USE CASE
Map the recruiting process
including all important
process steps (tasks)!
PROCESS ANALYSIS
Select the process steps that
can be standardized!
SELECT
Focus on the processes with
business relevance!
FOCUS
Translate the selected
process steps into
algorithms or out-of-the-box
tools!
TRANSLATE
Breaking down silos...the expedition
into HR's technological future will
not take place in a silo!
DANIEL MÜHLBAUER
HR MUST HAVE A SOLID
UNDERSTANDING OF HOW TO DESIGN
AND SCALE INTELLIGENT TECHNOLOGY
TECH MUST HAVE A SOLID
UNDERSTANDING OF THE
HUMAN DIMENSION OF WORK
THE MISSING LINK?
COMPUTER LINGUISTICS
EXPLAINABLE AI &
FAIRNESS
SOCIAL ROBOTICS &
AFFECTIVE COMPUTING
VIRTUAL & AUGMENTED
REALITY
THE BIG LEAPS?
HAPPY TO CONNECT!

AI in Recruiting: The Good, the Bad, and the Ugly

  • 1.
    AI in Recruiting Dr. D a n i e l M ü h l b a u e r T H E G O O D , T H E B A D A N D T H E U G L Y
  • 2.
    Knowing vs. Doing currentlyuse or implement AI projects in HR. think AI is a highly relevant technology. 79% of firms 8% of firms AI-study 2020 by Deloitte
  • 3.
  • 4.
    Robot-Coworker = popular? ofthe surveyed people are optimistic and excited about robot-coworkers (n=8000) 53% 64% 36% AI@WorkReport by Oracle of the surveyed people would trust a robot more than their manager (n=8000) of the surveyed people think that robots are less biased than managers (n=8000)
  • 5.
    MALE APPLICANTS RECEIVE SIGNIFICANTLY MORERECRUITER CALLBACKS THAN FEMALE APPLICANTS. (Rivera/Tilcsik, 2016) MALE APPLICANTS ARE PREFERRED FOR "MALE OCCUPATIONS", BUT WOMEN ARE NOT PREFERRED FOR "FEMALE OCCUPATIONS". (Koch et al., 2015) NO REVERSAL OF DISCRIMINATION IN RECRUITING AFRICAN- AMERICAN APPLICANTS SINCE 1989. (Quillian et al., 2019) FEMALE APPLICANTS OVER 40 RECEIVE SIGNIFICANTLY FEWER RECRUITER CALLBACKS. (Carlsson et. al. 2019) APPLICANTS WITH A MIGRATION BACKGROUND RECEIVE SIGNIFICANTLY FEWER RECRUITER CALLBACKS. (Moa, 2017) FEMALE RECRUITERS PREFER WOMEN. MALE RECRUITERS DO NOT PREFER MEN. (Carlsson/Eriksson, 2019) WOMEN GENERALLY RECEIVE FEWER RECRUITER CALLBACKS, ESPECIALLY IF THEY HAVE CHILDREN. (Gonzales et al., 2019) SHOULD WE PERHAPS GIVE INTELLIGENT HR TECHNOLOGY A CHANCE AFTER ALL?
  • 6.
  • 7.
    The solution tothe hiring puzzle is finding your firm’s happy place between HR Tech and HR Touch! DANIEL MÜHLBAUER
  • 8.
    THE GOOD RECRUITING ANALYTICS& PROCESS MINING Intelligent technology analyzes your recruiting funnel and identifies errors, circularities, inefficiencies of your recruiting. INTELLIGENT STRING- BUILDER Intelligent technology analyzes your job posting and suggests the best Boolean string for active sourcing. JOB-ADS w/ COMMUTE- ANALYSIS For each job posting, candidates receive an analysis of the work route including seasonal weather effects, means of transport and rush-hours. TEXT-MINING OF COMPANY INFOS An intelligent technology analyzes all publicly available company information and derives various indicators of employer quality from the language used. Good AI-based REC technology focuses specific HR tasks and enables a linkage of artificial and human intelligence!
  • 9.
    THE GOOD SKILL-BASED MATCHING Intelligenttechnology compares requirements and qualifications at the ski level and predicts a match quality. AUTOMATED APTITUDE TESTING Intelligent technology combines validated methods of aptitude diagnostics to predict aptitude in real time. PERSONALIZED CAREERPAGES Intelligent technology modifies career site content based on target group analysis and typical click behavior. AVATARS FOR RECRUITING TRAINING Intelligent virtual avatars enable the realistic simulation of interviews for a variant-rich training of recruiting trainees. Good AI-based REC technology focuses specific HR tasks and enables a linkage of artificial and human intelligence!
  • 10.
    THE BAD EMOTION RECOGNITION Emotionrecognition is a very serious research direction, but its results have not yet reached market maturity. FIT-PREDICITION BASED ON X The fit between people, jobs, teams, companies or cultures are very complex. The data models are often not reliable. AD TARGETING ON SOCIAL MEDIA Job posting algorithms are not transparent enough. They can subtly discriminate. Poor AI-based REC technology focuses on contexts that are too complex, are opaque, or are still in validation. As a result, their use should not yet be full- scale, but in pilots or validation studies.
  • 11.
    THE UGLY INVALID SELECTIONMETHODS Poor selectionprocessescannot be salvaged with AI use. Examples: Graphology, MBTI, DISG, profiling. DISCRIMINATION IN VIDEO RECRUITMENT Algorithms for image and video processing often work significantly worse for non-light-skinned people: PERSONALITY-ANALYSES -> PERFORMANCE Personality is not a stable predictor of subsequent job performance. AI- based technology does nothing to change that. Bad AI-based REC technology is based on dubious or even discriminatory procedures whose automation reinforces systematic disadvantage! FACE-ANALYSES -> ATTITUDES & BEHAVIOR Analyses of micro-expressionsto infer any traits or emotions of a person are not valid.
  • 12.
    YOUR USE CASE Mapthe recruiting process including all important process steps (tasks)! PROCESS ANALYSIS Select the process steps that can be standardized! SELECT Focus on the processes with business relevance! FOCUS Translate the selected process steps into algorithms or out-of-the-box tools! TRANSLATE
  • 13.
    Breaking down silos...theexpedition into HR's technological future will not take place in a silo! DANIEL MÜHLBAUER
  • 14.
    HR MUST HAVEA SOLID UNDERSTANDING OF HOW TO DESIGN AND SCALE INTELLIGENT TECHNOLOGY TECH MUST HAVE A SOLID UNDERSTANDING OF THE HUMAN DIMENSION OF WORK THE MISSING LINK?
  • 15.
    COMPUTER LINGUISTICS EXPLAINABLE AI& FAIRNESS SOCIAL ROBOTICS & AFFECTIVE COMPUTING VIRTUAL & AUGMENTED REALITY THE BIG LEAPS?
  • 16.