IEEE Computational Intelligence Magazine - May 2021 - 49

behavior, and counterproductive work behavior [21], [43]-
[46]. The surveys are designed to capture behaviors associated
with achieving organizational goals [47]-[49].
1)	Task Performance: We measure task performance using two
variables: In-Role Behavior (IRB) [29] and Individual Task
Proficiency (ITP) [30] using instruments validated with
significant samples [29], [30]. The former measures an
individual's perception of her job performance based on
completion of tasks associated with that individual's position. The latter measures the individual's perception of
how frequently she completed her core job tasks, completed these tasks well, and verified that these tasks were
completed well.
2)	Organization Citizenship Behavior: Organizational citizenship behavior was assessed using the Organizational
Citizenship Behavior Checklist (OCB-C; [31]). OCBs
are optional actions that are not rewarded by a worker's
organization. OCB-C is a validated instrument as
described in [31].
3)	Counterproductive Work Behavior: (CWBs) are actions taken
by employees that intentionally harm either the organization or individuals within the organization [50]. To measure CWB, we use the Interpersonal and Organizational
Deviance (IOD) scale [32]. The 19-item instrument is broken into two major categories of items: (1) Interpersonal
Deviance (7 items) and (2) Organizational Deviance (12
items). Each item has a seven-point frequency score: 1
(never) to 7 (daily). In our predictions, we consider each
major category as a separate variable. IOD was validated as
described in [32].

C. Health and Physical Variables

1)	Alcohol Consumption: We use the Alcohol Use Disorders
Identification Test (AUDIT), which was developed by the
World Health Organization (WHO; [37]). The effects of
alcohol consumption on job performance and other
areas of people's lives are well -documented [62]-[67].
AUDIT validity has been widely confirmed, e.g., [68]-
[70].
2)	Physical Activity: We use the International Physical Activity
Questionnaire (IPAQ; [39]). Physical activity affects not only
physical health but also mental well-being and job performance [71]-[74]. The test-retest reliability for the IPAQ
questionnaires is reported in [39].
3)	Sleep: We use the Pittsburgh Sleep Quality Index (PSQI;
[40]). Sleep is critical to incorporate because poor sleep
directly impacts job performance [75], cognitive ability, and
mental health [76], [77]. PSQI has both good internal reliability and good test-retest reliability [40].
4)	Tobacco Use: We used IARPA's modified version of the
Global Adult Tobacco Survey (GATS) from the World
Health Organization (WHO, [38]), which focuses only on
the individual's consumption by considering three items:
whether the participant is a current smoker, if they use
tobacco daily, and the quantity used in the past week. We
predict the last item only. Tobacco use is associated with
stress, negative emotionality, lower agreeableness, [67],
[78]-[82], and work performance [83]. GATS was
reviewed and approved by the GATS Questionnaire
Review Committee of the WHO.
III. Related Work

B. Psychological Constructs

We focus on four psychological constructs: cognitive ability,
personality, affect, and anxiety.
1)	Cognitive Ability: We use the Shipley Institute of Living
Scales 2 (Abstraction and Vocabulary sub-tests) [33] to measure fluid and crystallized intelligence respectively [51], [52].
See [53] for a study on the relation of cognitive ability and
job performance. Shipley 2 has high reliability and internal
consistency [33].
2)	Personality: Personality was measured in the initial ground
truth battery via the Big Five Inventory-2 (BFI-2; [34]). The
BRI-2 traits are: Extraversion, Agreeableness, Conscientiousness, Neuroticism, and Open-Mindedness. Each of the Big
Five Personality Traits has varying levels of association to job
performance [11], [12], [14], [15], [54]-[56]. BFI-2 was validated with four datasets [34].
3)	Affect: We use the Positive and Negative Affect ScheduleExpanded Form (PANAS-X; [35]). Affect variation is a key
indicator of a person's mental health and is critical for job
performance and other job behaviors [57-60]. PANAS-X
is shown to be a reliable instrument [61].
4)	Anxiety: We use the State-Trait Anxiety Inventory (STAI;
[36]). Anxiety is another key indicator of a person's mental
health. STAI was validated by [36].

We discuss related work for each of the categories of dependent variables in our study.
Psychological Variables. Empirical evidence shows that cognitive ability is related to personality traits. For example, personality (introversion) and abstraction are shown to be associated
with intellectual curiosity, which along with other traits can
partially explain crystallized intelligence [86].
Personality has been predicted by an individual's behavior,
ranging from, e.g., a person's interactions with computers
[91]-[93] to social pressures [102]. The field of automated
TABLE II Literature per variable group and predictor types.
VARIABLE GROUP

SENSOR/SOCIAL

ATTRIBUTES & TRAITS

Job Performance

[17] [18] [42]

[9-15, 84-87]

Cognitive Ability

[88, 89]

[90]

Personality

[18, 91-101]

[86, 102]

Affect

[103-109]

[106]

Anxiety

[104, 110, 111]

[36, 112, 113]

Alcohol

[114, 115]

[116]

Tobacco

[117]

[118]

Physical Activity

[119, 120]

[121]

Sleep

[122-125]

[125]

MAY 2021 | IEEE COMPUTATIONAL INTELLIGENCE MAGAZINE

49
IEEE Computational Intelligence Magazine - May 2021

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2021
