Examining the Implications of Automaticity Theory in the Construction Industry

Abstract

This study investigated the development of automaticity during repetitive construction activities. Twenty- eight subjects were recruited to participate in a total of 22 trials of simulated roofing installations for one month in a laboratory. The performance metric considered in this research for feature-based diagnosis of automaticity development was the roofing task accuracy. The results revealed that the roofing task accuracy significantly improved with practice as the trial days progressed. Given that practitioners are interested in training workers to achieve automaticity to increase their productivity and multi-tasking skills, the results of this study provide measures to test training effectiveness and the extent to which workers have developed automaticity. Also, by better understanding the model of humans, this study’s results will help improve human-AI teaming as the AI will better understand the cognitive state of its human counterpart and can adapt to him/her more accurately.

Keywords

Automaticity Theory Multi-tasking Performance Assessment Construction industry

Introduction

Automaticity is a critical skill component (Logan 1985) that is achieved with repeated practice or learning and enables one to perform a task automatically, unconsciously as a reflex, or with little to no attention (Stefanidis et al., 2007). It is similar to executing a task on autopilot. It is a situation where attentional demand/requirement increasingly diminishes as expertise increases in the execution of a task. It improves how quickly, smoothly, and effectively a skill is executed (Williams & Ford, 2008). Speaking and driving are typical instances of automaticity. Expert drivers, for instance, may have a technical discussion or prepare a speech while driving, even though it may be dangerous. Additionally, some drivers lose track of the specifics of the route after traveling safe distances. These are all instances when the tasks have been somewhat automatized due to repeated practice or skill development.

Researchers from a variety of fields, including art (Kozbelt and Seeley, 2007) and surgery (Reznick & MacRae, 2006), have noted the significance of extended practice and automaticity in converting laborious unskilled slow performance into effortless fast expertise (Shiffrin & Schneider, 1977; Reznick & MacRae, 2006; Charlton and Starkey, 2013). The crucial role that automaticity plays in expert performance has long been recognized by psychologists (James, 1890). Therefore, some benefits are associated with automaticity or automatic performance of tasks, such as efficiency, speed, and effortlessness. On the other hand, a limited number of studies have examined the empirical evidence demonstrating that highly automated activities could negatively affect expert performance (Toner et al., 2015) and may lead to accidents (Charlton and Starkey, 2013). For instance, Toner et al. (2015) examined a wide range of empirical evidence that highlighted various flaws that occasionally may be observed with respect to highly automated performances. These problems can be linked to automatic performances characterized by overfamiliarity (Charlton and Starkey, 2013) and a decline in attention, awareness, and consciousness (Saling & Phillips, 2007, p. 2;

Toner et al., 2015). This is because as automaticity develops, attention becomes increasingly reduced, making it more likely that changes in a familiar environment will go undetected and that dangerous situations won't be recognized, increasing the likelihood that an accident will occur. Considering this, several epidemiological researchers have long recognized that accidents are more likely to happen on familiar roads (Durand, 1980; Chen, et al., 2005; Charlton and Starkey, 2013). Additionally, based on these findings, researchers in different fields have examined the connection between automaticity and attention to explain the underlying causes of accidents and hazard mitigation techniques (Toner et al., 2015). These scholars have presented findings that suggest that automaticity could impair safety.

Even though studies on the concept of automaticity and its connection to accident involvement have been conducted in other domains, such as driving (Charlton and Starkey, 2013; Panek et al., 2015), little is known about how automaticity affects the safety of construction workers.

To fill this knowledge gap, this current study examines methods that indicate the development of automaticity on construction sites and its effects on construction safety performance. The specific objective of the present study is to investigate the development of automaticity during repeated construction tasks.

To achieve this objective, a roofing experiment involving the installation of shingles on a simulated roof model was designed, and participants were required to complete 22 trials of simulated roofing installations. To enable experimental manipulations, the 22 trials in this investigation were further divided into four trial days (i.e., days 1, 2, 3, and 4). To answer the study questions, the data will be analyzed using various statistical methods, such as pairwise comparisons and repeated measures analysis of variance.

This study advances academic knowledge by demonstrating the development of automaticity in the execution of construction tasks. Safety managers may utilize this study's results to better understand how automaticity affects construction safety. The findings and measurements from this research can be used to test the efficacy of training and the degree to which workers have developed automaticity, which is important to practitioners because it helps to increase workers' productivity and multitasking abilities.

Background

Diagnosis of Automaticity

According to Logan (1980) and Moors and De Houwer (2006), researchers in the mechanics of thoughts have reportedly long been interested in the identification and explanation of a clear distinction between automatic and control processes (Laberge, 1973; Posner & Snyder, 1975a; Logan, 1979; Reingold & Sheridan, 2011; Wu, 2020). They are interested in identifying particular criteria that, if met, could help determine/diagnose when a nonautomatic process (i.e., an attentionally controlled process), turns automatic (Shiffrin & Schneider, 1977; Logan, 1979; Stefanidis et al., 2012; Wu, 2020). The majority of researchers have identified automaticity in performances or processes by looking for characteristics like attentionally demanding, unintentional uncontrolled/uncontrollable, unconscious, purely stimulus- driven, autonomous goal independent, efficient, and fast (Moors and De Houwer, 2006). This is referred to as a feature- based diagnosis.

Performance Assessment Metrics for Feature-Based Diagnosis

Part of what distinguishes skilled performers (experts) from novices is their capacity to participate in specific tasks without excessively utilizing their attentional resources to produce quick, precise, and effective results (Stefanidis et al., 2012). This capacity was first referred to as automaticity by psychologists (Shiffrin & Schneider, 1977; Logan, 1988). Hence, observing the presence of intrinsic characteristics like speed, error (or accuracy), and attentional demand allows one to diagnose, quantify, or analyze this capacity (automaticity). These characteristics include traditional performance evaluation metrics like time (or duration or speed) and error (Stefanidis et al., 2012). Others may be referred to as nontraditional performance evaluation metrics, such as secondary task performance metrics that measure spare attentional capacity (Stefanidis et al., 2012). Most simulation curricula currently in use evaluate and give feedback on performance using traditional metrics of error (or accuracy) and time. (Stefanidis et al., 2012).

Knowledge Gap : There are no metrics in construction for measuring/diagnosing automaticity, and hence the performance of these metrics in construction has not been tested, and their applicability has not been evaluated. To bridge this gap in knowledge, this current study tested automatic performance measures within the construction industry setting. This study adopted a traditional indicator of accuracy or efficiency in diagnosing the development of automaticity during repetitive construction activities.

Point Of Departure

Construction activities involve highly attentionally demanding tasks. Additionally, due to the dynamic and dangerous nature of construction sites, construction workers must divide their attention to both tasks and their environment—in the interest of safely completing their task—and maintain good situation awareness during task execution (Hasanzadeh et al. 2017, 2018). Furthermore, since some construction activities, such as shingle installations, involve repetitive tasks, some aspects may become automated with practice (Ranney, 1994). Previous investigations on automatic behaviors in various domains have suggested that this state is characterized by a lack of attention (LaBerge & Samuels, 1974; Posner & Snyder, 1975a, 1975b; Shiffrin & Schneider, 1977; Logan, 1979). Despite the crucial role of attention in hazard identification (Hasanzadeh et al., 2017) and construction safety, little is known regarding the development of automaticity in repetitive construction tasks.

This study aims to address this research gap by investigating methods that are indicative of the development of automaticity on construction sites and its implications on construction safety. To accomplish this aim, the specific objective to be achieved in this study involves examining the development of automaticity during repetitive construction activities. To test whether participants developed automaticity during repetitive construction activities, task accuracy was used in the simulated experiment.

The following null hypothesis was tested to examine the differences in automatic performance measures for diagnosing automaticity development during repetitive construction activities:

Null Hypothesis #1: There is no significant difference between the means of the roofing task accuracies (one of the features of automaticity development) measured during repetitive construction activities (i.e., measured across trial days).

Hence the combined null hypothesis is that the mean automatic performance measures will not vary during repetitive construction activities (i.e., across trials).

Research Method

Experimental Design

Following the research protocol established in previous studies (Stefanidis et al., 2007, 2012), this experiment involved a roofing installation task. To determine if automaticity was developed during repetitive construction activities, the performance of participants for each of these activities was collected and compared across trials. The details of these tasks are provided here (see Figure 1).

Figure 1.

Roofing experimental design.

Roofing Installation Task

This experiment’s task involved installing shingles on a simulated roof model. Twenty-eight participants were asked to stand on a low-sloped simulated roof model 4ft wide, 6ft long, and 3ft high while installing 17 pieces of 25 ft² shingles on a roof model in a construction safety laboratory as presented in Figure 1. The participants gently hammered the Class A fire rating shingles into place on the rooftop, and the shingles were held in place with Xfasten Double-Sided Woodworking Tape that was already attached to the back of the shingles. The performance of participants in the roofing activity was obtained by measuring the accuracy of the shingle installation.

Roofing Task Accuracy

To measure the accuracy of shingle installation, the researcher obtained screenshots of the installed shingles at the end of the experiment. The shingles installation accuracy scores were assessed by five (two roofing project managers, two professional roofers, and the researcher) construction professionals (with more than 10 years of work experience in the construction industry). Five expert raters were used in this research to minimize the possibility of personal bias (see Figure 1). There were 616 (22 trials for 28 participants) graded roofing installation images. Each of the experts gave their expert ratings of the 616 completed roofing installation images with respect to how it should be installed in practice. Each rater took about 7 hours to grade every single installation; hence, breaks of 30 minutes were taken at every two-hour interval, and the 616 images were graded at an average of six effective hours (minus the breaks). Since the experts utilized a percentage scale, their raw ratings ranged from 0 to 100. These raw ratings were further normalized to guarantee that all accuracy ratings (z-scores) were standardized on the same scale. The 616 accuracy scores of each of the professional raters were standardized based on the mean and standard deviation of the scores of each rater. Therefore, the standard deviation of all the scores per expert was 1, and the mean was 0. The roofing task accuracy score for each participant in a session was calculated as the average of the five standardized accuracy scores from the five expert raters.

Data Collection

Before the experiment, the participants were given a thorough explanation of the equipment setup, experimental protocols, specific roofing activities, and other important information. To minimize the risk of injury during the task, participants were provided with securely fastened Personal Protective Equipment (PPE) including safety harness, knee guards, and safety gloves. A Sony Zeiss Vario-Sonnar T* video camera was used to record the experiment for further examination and video analysis. The factors considered in this study were the automatic performance measures (such as roofing task accuracy scores) and trial day. These factors were divided into two categories: dependent variables (i.e., performance indicators such as the accuracy of the roofing task) and independent variables (i.e., trial day).

Participants

This study conducted this experiment among 28 recruited university students aged 18-33 years (M = 24 years, SD = 4 years) who were tasked with roofing activities. The participants involved 23 male and 5 female students recruited from the Civil and Infrastructure Engineering department at George Mason University, Fairfax, Virginia, and on average, had 1 year of work experience (Range = 0-9 years, SD = 2 years). The subjects regularly participated in about 22 simulated roofing installations (at an average of 10 minutes per installation) for a period of one month and were rewarded with a gift voucher for completing the entire experimental sessions. Twenty-eight participants and twenty-two trials were used in this research in line with previous/similar studies in other fields (Stefanidis et al., 2007; Charlton & Starkey, 2011, 2013).

Data Analysis

To achieve the objective of this study, the data collected on the roofing task accuracy for the 22 trials of the experiment were analyzed using pairwise comparisons and repeated measures analysis of variance (RM-ANOVA) to detect any overall differences between trial means, thereby ascertaining whether automaticity was developed during the repetitive roofing activity.

RM-ANOVA Assumptions

Valid use of the RM-ANOVA procedure is contingent on certain assumptions. As a result, researchers examine whether these assumptions/conditions have been met to confirm the result’s credibility, validity, and reliability (Hahs-Vaughn & Lomax, 2013). The assumptions considered in this research include independence, normality, sphericity, continuous dependent variable, categorical independent variable, and no significant outliers. These required assumptions were tested in this study with respect to the RM-ANOVA to verify the credibility of the results. In summary, the above assumptions were met, and analysis techniques (pairwise comparisons and RM-ANOVA) were employed to test the research hypothesis of whether automaticity is developed during repetitive construction activities. A p-value less than 0.05 was considered significant in this experiment.

Automaticity Development Test Results

To test the development of automaticity during repetitive construction activities, this study considered the roofing task accuracy as the automatic performance measure. To facilitate experimental manipulations, the 22 trials in this experiment were further divided into trial days, where day 1 corresponds to trials 1 through 6, day 2 corresponds to trials 7 through 12, day 3 to trials 13 through 18, and day 4 to trials 19 through 22. The analysis results of the roofing task accuracy across the four days are presented below.

The central hypothesis of this analysis is that mean automatic performance measures (i.e., roofing task accuracy) will vary significantly over the entire period of the experiment, which spans from day 1 to day 4 (i.e., trial 1 to trial 22). Hence the combined null hypothesis is that the mean automatic performance measures will not vary across trials.

Roofing Task Accuracy Results

The null hypothesis of this section of the analysis results is stated in detail as follows:

Null Hypothesis #2: There is no significant difference between the means of the roofing task accuracies (that measures one of the features of automaticity development) measured across trial days.

This hypothesis was tested using the RM-ANOVA. This study also examined possible pairwise differences in means using Bonferroni-adjusted pairwise comparisons (i.e., paired t-tests with Bonferroni correction to minimize the risk of a type I error).

Repeated Measures ANOVA

The repeated measurements of the participants' automatic performance measure (dependent variable i.e., roofing task accuracy) across phases of the trial (independent variable) were investigated in this analysis using the RM-ANOVA technique in SPSS. The results of the statistical analysis (as obtained in the multivariate and the univariate tests) indicated that there was a main effect of the trial day on roofing task accuracies with a p-value less than 0.05 in the outputs.

According to the RM-ANOVA test, this study found statistical significance and hence, rejected the combined null hypothesis that says that there is no significant difference between the means of the roofing task accuracies measured across trial days (p-value < 0.05). Therefore, there were significant differences between the mean roofing task accuracies (i.e., mean automatic performance measures) over the repeated measurements (p-value < 0.05). In other words, the mean roofing task accuracy or quality of work significantly improved with practice as the trial days progressed. Since task accuracy (automatic performance measure) provides an index for evaluating feature-based improvements that are synonymous with the development of automaticity (Stefanidis et al., 2012), this current study draws on this evidence to argue that automaticity is developed during repetitive construction activities.

Pairwise comparisons

Since this study discovered that there was a statistically significant effect of trial day on the automatic performance (i.e., mean roofing task accuracy in this case) measure (p < 0.001), a post-hoc test/analysis involving pairwise comparisons was implemented to identify the exact trial day(s) where the effect(s) or difference(s) exist(s). The Bonferroni- adjusted pairwise comparisons, which are paired t-tests with p- values adjusted for multiple comparisons, show that the mean roofing task accuracy varied (or improved) significantly from day 1 to every other day (i.e., days 2, 3, and 4). Also, the mean roofing task accuracy varied (or improved) significantly from every other day (i.e., days 1, 2, and 3) to day 4, with day 4 recording the highest mean roofing task accuracy and day 1 recording the least mean task accuracy. However, the mean roofing task accuracy did not vary (or improve) significantly from day 2 to day 3. This means that over time, the quality of the shingle installation task (roofing task) of the participants significantly improved during the repeated construction activity, and this could be an indication of the development of automatic behavior since increased quality of work or task performance is one of the attributes of the development of automaticity (Stefanidis et al., 2012). However, this improvement is evident (or significant) when compared between the first day (day 1) and every other day or between every other day and the last day (day 4).

Discussion

Automaticity development in construction

In summary, an RM-ANOVA was employed in this study to investigate whether there were statistically significant differences (i.e., notable improvements) in automatic performance measurements (e.g., accuracy) over the course of the execution of a roofing experiment that was repeatedly executed for four days within one month. There were no significant outliers, and the automatic performance data collected in this study were normally distributed for each of the four trial days (i.e., days 1, 2, 3, and 4) as evaluated by boxplot and Shapiro-Wilk’s test for the roofing task accuracy (p-values = 0.235, 0.928, 0.101, 0.698). The assumption of sphericity was not violated for roofing task accuracy as evaluated by Mauchly's Test of Sphericity, p-values = 0.054.

Thereafter, the following conclusions were inferred from the univariate (within-subjects effects) RM-ANOVA test. There was a statistically significant effect of trial day on the mean roofing task accuracy, F(3, 81) = 13.71, p < 0.001, η ² = 0.34.

Additionally, the following conclusions were inferred from the multivariate test. There was a statistically significant effect of trial day on the mean roofing task accuracy, Wilks' Lambda = 0.51, F(3, 25) = 8.14, p < 0.001, η ² = 0.49. In other words, there was a main effect of the trial day on the automatic performance measures examined in this study with p-value < 0.05. The performance measures recorded a large effect size (η ² > 0.14) according to Cohen's (1988) benchmarks for small (0.01), medium (0.06), and large (0.14) effects using partial eta-squared (η ²). The automatic performance measure examined in this research significantly improved over time. The post-hoc analysis with a Bonferroni adjustment revealed that the mean improvement in the automatic performance measure (i.e., accuracy) was significant from the first day (day 1) to every other day (i.e., days 2, 3, and 4), and from every other day (i.e., days 1, 2, and 3) to the last day (i.e., day 4). However, the improvement from day 2 to day 3 was not significant.

Generally, the result of this study revealed that the automatic performance measure (accuracy) exhibited a significant improvement from day 1 (novice stage) to every other day (i.e., days 2, 3, and 4). Hence, the performance in the novice stage is significantly different from every other stage, which is an indication that an automaticity feature significantly changed/improved, and this could be an indication of the development of automaticity.

Significance and Conclusions

This research is the first effort to apply automaticity theory from the field of psychology to the construction industry. Although this research in the construction industry is still in its early phases, the results are encouraging. Both academia and practice can benefit significantly from this study. Concerning contributions to academia, this research provides evidence for the development of automaticity in the performance of construction tasks. Regarding contributions to the industry, the results of this study provide metrics to test the effectiveness of training and the degree to which workers have developed automaticity, which is relevant to the construction industry because practitioners are interested in training workers to achieve automaticity to increase productivity and multitasking skills. Also, the results will help safety managers better understand the implications of automaticity on hazard identification among construction workers and design appropriate risk mitigation strategies. Finally, the findings of this study will help to enhance human-AI teaming since the AI will be able to adapt to its human counterpart more correctly due to having a better knowledge of the human model.

Footnotes

Acknowledgements

The National Science Foundation is thanked for supporting the research reported in this paper through the Future of Work at the Human-Technology Frontier (FW-HTF) program. This paper was based on work supported by the National Science Foundation under Grant No. 2310210. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

ORCID iDs

Ikechukwu Sylvester Onuchukwu

Behzad Esmaeili

References

Charlton

S. G.

Starkey

N. J.

(2013). Driving on familiar roads: Automaticity and inattention blindness. Transportation Research Part F: Traffic Psychology and Behaviour, 19, 121-133.

Charness

Reingold

E. M.

Pomplun

Stampe

D. M.

(2001). The perceptual aspect of skilled performance in chess: Evidence from eye movements. Memory & Cognition, 29(8), 1146-1152.

Chen

I. G.

Durbin

D. R.

Elliott

M. R.

Kallan

M. J.

Winston

F. K.

(2005). Trip characteristics of vehicle crashes involving child passengers. Injury Prevention, 11(4), 219-224.

Durand

(1980). An Analysis of Accident Location in Relation to Area of Residence (No. AIRAC-A80-4).

Hasanzadeh

Esmaeili

Dodd

M. D.

(2017). Impact of construction workers’ hazard identification skills on their visual attention. Journal of Construction Engineering and Management, 143(10), 04017070.

Hasanzadeh

Esmaeili

Dodd

M. D.

(2018). Examining the relationship between construction workers’ visual attention and situation awareness under fall and tripping hazard conditions: Using mobile eye tracking. Journal of construction engineering and management, 144(7).

James

(1890). The principles of psychology 1963 New York Holt. Rinehart & Winston (Original work published 1890).

Kozbelt

Seeley

W. P.

(2007). Integrating art historical, psychological, and neuroscientific explanations of artists' advantages in drawing and perception. Psychology of Aesthetics, Creativity, and the Arts, 1(2), 80.

LaBerge

(1973). Attention and the measurement of perceptual learning. Memory & Cognition, 1(3), 268-276.

10.

LaBerge

Samuels

S. J.

(1974). Toward a theory of automatic information processing in reading. Cognitive Psychology, 6(2), 293-323.

11.

Logan

G. D.

(1979). On the use of a concurrent memory load to measure attention and automaticity. Journal of Experimental Psychology: Human Perception and Performance, 5(2), 189-207

12.

Logan

G. D.

(1985). Skill and automaticity: Relations, implications, and future directions. Canadian Journal of Psychology/Revue Canadienne de Psychologie, 39(2), 367.

13.

Logan

G. D.

(1988). Automaticity, resources, and memory: Theoretical controversies and practical implications. Human Factors, 30(5), 583-598.

14.

Moors

De Houwer

(2006). Automaticity: a theoretical and conceptual analysis. Psychological bulletin, 132(2), 297.

15.

Neisser

Becklen

(1975). Selective looking: Attending to visually specified events. Cognitive Psychology, 7(4), 480-494.

16.

Panek

E. T.

Bayer

J. B.

Dal Cin

Campbell

S. W.

(2015). Automaticity, mindfulness, and self-control as predictors of dangerous texting behavior. Mobile Media & Communication, 3(3), 383-400.

17.

Posner

M. I.

Snyder

C. R. R.

(1975a). Attention and cognitive control. In Solso

R. L.

(Ed.), Information processing and cognition: The Loyola symposium (pp. 153–175). Hillsdale, NJ, 55-85. Erlbaum.

18.

Posner

M. I.

Snyder

C. R. R.

(1975b). Facilitation and inhibition in the processing of signals. In Rabbit

P. M. A.

Dornic

(Eds.), Attention and performance V (pp. 669 – 682). New York: Academic Press

19.

Ranney

T. A.

(1994). Models of driving behavior: a review of their evolution. Accident Analysis & Prevention, 26(6), 733-750.

20.

Reingold

E. M.

Charness

(2005). Perception in chess: Evidence from eye movements. In Underwood

(ed.) Cognitive processes in eye guidance, 325-354. Oxford: Oxford University Press.

21.

Reingold

E. M.

Sheridan

(2011). Eye movements and visual expertise in chess and medicine. In Liversedge

S. P.

Gilchrist

I. D.

Everling

(Eds.) The Oxford Handbook of eye movements (pp. 523–550): Oxford University Press

22.

Reznick

R. K.

MacRae

(2006). Teaching surgical skills—changes in the wind. New England Journal of Medicine, 355(25), 2664-2669.

23.

Saling

L. L.

Phillips

J. G.

(2007). Automatic behaviour: efficient not mindless. Brain research bulletin, 73(1-3), 1-20.

24.

Shiffrin

R. M.

Schneider

(1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending and a general theory. Psychological Review, 84(2), 127-190.

25.

Simons

D. J.

Chabris

C. F.

(1999). Gorillas in our midst: Sustained inattentional blindness for dynamic events. Perception, 28(9), 1059-1074.

26.

Stefanidis

Scerbo

M. W.

Korndorffer

J. R.

Jr Scott

D. J.

(2007). Redefining simulator proficiency using automaticity theory. The American Journal of Surgery, 193(4), 502-506.

27.

Stefanidis

Scerbo

M. W.

Montero

P. N.

Acker

C. E.

Smith

W. D.

(2012). Simulator training to automaticity leads to improved skill transfer compared with traditional proficiency-based training: a randomized controlled trial. Annals of Surgery, 255(1), 30-37.

28.

Toner

Montero

B. G.

Moran

(2015). The perils of automaticity. Review of General Psychology, 19(4), 431-442.

29.

Williams

A. M.

Ford

P. R.

(2008). Expertise and expert performance in sport. International Review of Sport and Exercise Psychology, 1(1), 4-18.

30.

(2020). Automaticity, control, and attention in skill. In The Routledge Handbook of Philosophy of Skill And Expertise (pp. 207-218). Routledge.