Computational Model for Changing Sedentary Behavior through

Cognitive Beliefs and Introspective Body-feelings

Fawad Taj

1,2 a

, Nimat Ullah

1,3 b

and Michel Klein

Social AI group, Dept. of Computer Science, VU Amsterdam, Amsterdam, The Netherlands

Dept. of Computer Science, University of Swabi, Swabi, Pakistan

Dept. of Computer Science, FATA University, TSD Dara, Kohat, Pakistan

Keywords: Digital Health, Sedentary Behavior, Theory of Planned Behavior, Health Belief Model, Network Oriented

Modelling.

Abstract: Sedentary behavior has emerged as a serious risk factor for numerous health outcomes. However, little work

has been done to approach the problem through social-cognitive theories. In this study, a network model has

been proposed for sedentary behavior intervention based on Influential determinants from major social-

cognitive theories i.e., theory of planned behavior and health-belief model. Accounting for these determinants

means that we are influencing behavior with a peripheral route, for which we included the somatic markers

as a body-feelings in the model. An effective behavior change techniques from literature are used to affect

these determinants to change the sedentary behavior. The model has been mathematically represented and

simulated using a network-oriented modelling technique for an office employee.

1 INTRODUCTION

Sitting behaviour is characterized by any waking

behavior with an energy expenditure of ≤1.5

metabolic equivalents (METs). You can be sedentary

at work, at school, at home, when travelling or during

leisure time while watching television, studying or

working at a desk or computer. A person can do

enough physical activity to meet the guidelines and

still be considered sedentary if he/she spends a large

amount of his/her time sitting or lying down

(Weggemans et al., 2018). Moreover, low level or

moderate-to-vigorous level physical activity is not the

same as being sedentary for example, I cycle to the

office every day (which is Dutch culture) and then sit

at a computer for around 6-7 hours, so it is possible

for being highly sedentary and highly active at the

same time. Prolonged sitting has several adverse

health outcomes including increased risk of type 2

diabetes, higher risk of premature death and death

from cardiovascular disease (Australian Government

guidelines for sedentary behavior, 2019)

(Weggemans et al., 2018).

https://orcid.org/0000-0001-9049-1736

https://orcid.org/0000-0002-0592-8380

https://orcid.org/0000-0003-4119-1846

A number of theories and models from social and

behavioral sciences can assist us to make sense of

behaviour and the world around us. More specifically

for sedentary behavior, a number of ecological

model/theories are proposed, but they are seldomly

used. In a recent review (Huang, Benford, & Blake,

2019), 19 out of 63 digital interventions for sedentary

behavior are based on some theoretical grounds

(among them the theory of planned behavior is used

for 5 times and social cognitive theory for 4 times).

Whereas from digital technological prospective the

sedentary behavior intervention mostly uses mobile

apps and wearable sensors (Taj, Klein, & van

Halteren, 2019). Sedentary behavior interventions

usually follow ecological models that define

multifaceted determinants of the problem, including

individual, social, and environmental policy level

(Owen et al., 2011).

A shortcoming of the ecological model is that they

fail to acknowledge the role of psycho-social

variables in explaining sedentary behavior

(Prapavessis et al., 2015). On an individual level,

different characteristic like beliefs, motivation or

Taj, F., Ullah, N. and Klein, M.

Computational Model for Changing Sedentary Behavior through Cognitive Beliefs and Introspective Body-feelings.

DOI: 10.5220/0010247704430450

In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 5: HEALTHINF, pages 443-450

ISBN: 978-989-758-490-9

443

intention etc. can influence sedentary behavior. For

understanding these types of determinants, socio-

cognitive theories are the best options to be reached

out (Sallis, Owen, & Fisher, 2015). Sedentary

behavior in the workplace is high; 71–77% of

working hours are being spent sedentary (Scherer,

2005). It requires minimal effort or conscious

planning and is highly habitual. To change workplace

sedentary behavior, we need to target these

determinants using effective behavior change

strategies, which will also be discussed later in the

paper.

In this paper, we focus on a conceptual model that

considers psycho-social determinants to reason about

sedentary behavior and use different behavior change

techniques to break a sedentary behavior.

The aims of this research includes: Identifying

the key psycho-social determinants from different

health cognitive theories for sedentary behaviour.

Exploring the popular behavior change

strategies/techniques from literature to target these

determinants ans lastly, modeling the findings as a

computational network model and simulating an

office employee working scenario using the model.

In Section 2 of the paper, the background of the

constructs from different theories are given and

discusses the behavior change techniques (BCTs) that

can be used to influence these determinants. In

section 3, the conceptual and mathematical

representation of network-oriented model is

presented. Section 4 contains the scenario and the

simulation results. The paper has been concluded in

Section 5 with a brief insight into the future

directions.

2 BACKGROUND

This section provides the background for the model

we proposed. In the first part, theories and working of

its determinants/parameters are discussed with

linkage to sedentary behaviors. The second part of

this section discusses different behaviour change

techniques with its association to the determinants of

the theories.

2.1 Socio-psychological Determinants

Most of the health cognitive theories describe

possible relationships between the psycho-social

factors and sedentary behavior, but theory of planned

behavior (TPB) has been mostly used in this context

(Prapavessis et al., 2015). According to TPB, an

individual’s intention is the main determinant of

actual sedentary time. The intermediate determinants

of intention are attitude, subjective norms (SN), and

perceived behavioral control (PBC). Attitude

represents an individual’s evaluation of the perceived

benefits and cost of sitting, SN reflects a belief about

whether most people approve or disapprove an action,

and PBC refers to individual’s perception of their

ability to control the time they spend being sedentary

(Prapavessis et al., 2015, Ajzen, 2005).

Health belief model (HBM) is the theory mostly

used to identify the determinants which explain the

likelihood of engaging in health-promoting behavior.

Perceived outcomes and self-efficacy are the main

constructs in HBM. Similarly, from Social Cognitive

Theory (SCT), self-efficacy construct suggests a

setting of realistic and measurable goals to ensure

initial success and the outcome expectancies

construct would suggest highlighting the benefits of

reducing sedentary time e.g., reduced muscle stiffness

etc. (Owen et al., 2011). The application of social

cognitive theory for health behavior change has

focused predominantly on increasing self-efficacy,

for example, confidence in one’s own abilities

(Bandura, 1998). The perceived self-efficacy is

highly correlated with goal attainment, higher the

self-efficacy, higher the goals people set for

themselves (Bandura, 2004). There exists a fair

amount of cross-sectional studies that correlate social

cognitive constructs to workplace setting but the

association between social-cognitive factors and

sedentary behavior needs much more exploration

(Hadgraft et al., 2017).

Among different theoretical models, most of the

determinants are overlapping and most of the

researcher overload their studies with the dictum that

more is better (Bandura, 2004). The determinants

discussed above are basically the internal cognitive

beliefs of humans. We represented these determinants

as positive and negative beliefs, for example,

perceived benefit (HBM) and self-efficacy (HBM,

TPB, SCT) are the positive beliefs about the action.

Similarly, perceived severity and susceptibility to

disease/behavior (HBM) are the negative beliefs

about the outcomes. Moreover, subjective norms in

TPB corresponds to expected social outcomes for a

given behavior. Perceived behavioral control in TPB

overlaps with perceived self-efficacy in SCT

(Bandura, 2004).

2.2 Behavior Change Techniques

In any intervention, BCTs are an important active

ingredient that may explain study variation in-

effectiveness. Effective sedentary reduction

HEALTHINF 2021 - 14th International Conference on Health Informatics

444

intervention depends on understanding and reasoning

about what works and why (Michie et al., 2013). A

taxonomy is available which describes 93 discrete

behaviour change techniques that can be used in

interventions within any behavioural domain e.g.

providing information on health consequences,

setting goals, restructuring the physical environment.

Behaviour change techniques represent the

observable and irreducible intervention components

that serve to perform one or more functions (Michie

et al., 2013).

Developing intervention based on theories and

models suggest using the available systematically

verified BCTs to effectively target the determinants

noted in the above section. For example, from SCT

and TPB the use of self-efficacy construct suggests

the use of goal-setting e.g. setting a walking goal to

the corridor door after every 30 minutes of work, and

self-monitoring e.g. maintains sitting time record

book (Owen et al., 2011). BCT taxonomies are new

and are seldom reported in digital behavior change

interventions (Direito et al, 2016). It has been

observed that only 10 out of potential 93 BCTs were

present (mean of 2.42 BCTs were present in each app)

(Dunn, 2017). Table 1 shows the description of each

of the determinant with scale description, source

theory/model, and some of the effective BCT to target

these determinants, coded from recent BCTs

taxonomy (Michie et al., 2013). The best resource

available to find the linkage between BCT and the

above-mentioned determinants (mechanism of

action) is the human behaviour change project

(https://www.humanbehaviourchange.org/). In one of

the studies in this project, they published the

triangulating evidence of links made by authors in

published scientific studies and by expert consensus

(https://theoryandtechniquetool.humanbehaviourcha

nge.org/).

3 NETWORK-ORIENTED

MODEL

Network-oriented temporal-causal network

modelling approach (Treur, 2016) has been used for

modelling the above concepts. The temporal

dimension enables the modelling of cyclic causal

relations with exact timing. The reason for

representing it as a causal network lies in the fact that

most of the determinants are subjective beliefs and

they are causally linked like a network. There are two

types of beliefs presented in the model, the positive

and negative beliefs; self-efficacy and perceived

benefit are presented as positive beliefs about the

current sedentary behaviour and the perceived

susceptibility and severity are negative beliefs about

sedentary behaviour. The two types of actions i.e.,

sedentary behaviour (sitting) and non-sedentary

behaviour (walk) are inversely influenced by these

positive and negative beliefs.

The network-oriented models can be represented

in two ways i.e., graphical representation and

numerical representation. Fig. 1 shows the graphical

representation of the proposed model. In section 3.1,

the graphical representation is converted into a

numerical or mathematical representation.

In the model, a person sedentary behavior is

represented with the states i.e., preparation for action

(ps



) and execution of action (es



). Whereas, the

walking behavior is represented with states i.e.,

( ps



) and execution of action ( es



). The

determinants discussed in section 2 are represented by

different states in the network for example, perceived

susceptibility and severity are represented with state

name srs

.

, srs

.

respectively. The

scenario discussed in section 4, shows how perceived

susceptibility and severity of the action affects the

actions execution and how after the intervention,

efficacy and perceived benefits increases.

This shift in beliefs is model used the Damasio’s

somatic marker hypothesis, i.e., introspective

feelings. It plays a critical role in the ability to make

fast, rational decisions in complex and uncertain

situations (Damasio, 1998). The feeling actually

serves a kind of monitoring and helps in choosing the

best possible options for action. This feeling state is

affected by predictive as-if body loop, which gives a

sense of preview and valuing the action before it has

actually been performed.

3.1 Mathematical Representation

A network-model illustrated in fig. 1 involves states

that reflect actual world anomalies, and the arrows

indicate the causal connection between the two

entities. Important notions for each of the state and

connection are as follow:

• Connection(ω

X,Y

): represents the connection

value between the states (x,y). The value represents

the strength of causality and its value ranges between

[-1, 1].

Speed Factor (η

): How fast the state value going to

change with incoming causal impact.

Computational Model for Changing Sedentary Behavior through Cognitive Beliefs and Introspective Body-feelings

445

Table 1: Determinants with scale description and coded behavior change techniques.

Determinan

Scale Description BCT

Perceived

Severity

(HBM)

Negative beliefs

One's belief of how serious a condition and its

consequences are

E.g. How confident are you that long sitting can

cause serious chronic illnesses.

BCT: 5.1. Information about

health consequences

BCT: 9.2. Pros and cons

BCT: 10.1. Material incentive

(behaviour)

BCT: 10.10. Reward (outcome)

Perceived

Susceptibility

(HBM)

One's belief of the chances of getting a condition.

E.g. How confident are you that your health will

not be with long sitting.

BCT: 5.1. Information about

health consequences

BCT: 5.2 Salience of

consequences

Social norms

(SCT, TPB)

Perceived organization/social support for less

sitting at work.

E.g. My workplace environment has an open

choice to stand or move more at work.

BCT: 6.3. Information about

others’ approval

BCT: 6.2. Social comparison

Perceived

Benefits

(HBM)

Positive beliefs

One's belief in the benefit of the advised action to

reduce the risk or seriousness of the impact

E.g. How confident are you that small breaks after

every 30 minutes will help me avoiding chronic

disease.

4.1 Instruction on how to

perform behavior — how,

where, when

5.3 Information about social and

environmental consequences.

Self-Efficac

(HBM, SCT,

TPB)

Confidence in one's ability to act. Provide training,

guidance, and positive reinforcement

E.g. How confident would have been that you

could have stood up during the meeting.

BCT: 1.2. Problem solving

BCT: 8.7. Graded tasks

BCT: 4.1. Instruction on how to

perform behaviour

BCT: 6.1. Demonstration of the

ehaviou

Figure 1: The network model for sedentary behavior change. The red lines show the negative connections and black lines are

the positive connections.

HEALTHINF 2021 - 14th International Conference on Health Informatics

446

• Combination Function (c

(...)): is used to

combine the causal impact of multiple incoming

states. This approach provides a library of currently

40 combination functions for the aggregation of

multiple (incoming) causal impacts.

Impact

X,Y

= ω

X,Y

X(t)

(1)

Total aggregated impact on state Y at time t combined

by combination function.

aggimpact

(t) = c

(impact

X1,Y

, impact

X2,Y

impact

X3,Y

, …)

= c

(ω

X1,Y

(t), ω

X2,Y

(t), ω

X3,Y

X3(t), …)

(2)

The aggimpact

(t) will have upward or downward

effect at time point t, but how fast this change takes

place depends on the speed factor η

Y(t+

Δt) = Y(t) + η

[aggimpact

(t) – Y(t)] Δt

(3)

The following difference (eq.4) and differential

equation (eq.5) can be obtained for state Y:

Y(t+

Δt) = Y(t) + η

[cY(ω

X1,Y

X1(t),

X2,Y

X2(t), ω

X3,Y

X3(t), …) – Y(t)]Δt

(4)

dY(t)/dt = ηY [c

(ω

X1,Y

X1(t), ω

X2,Y

X2(t),

X3,Y

X3(t), …) – Y(t)]

(5)

3.2 Parameters Formalization

The states represented in the model are cognitive

states having continuous one-dimensional value e.g.

sad vs happy. The causal connections shown by black

and red arrow shows the positive and negative

influence of one state on other, respectively. We have

scaled all type of values in the range of [-1,1] and

simulation time (t) is set for 200 steps.

The parameters required to define the network

models are: initial value, connection value,

combination function, and speed factor for each state.

The initial value for each of the state are set according

to the situation e.g., in simulation below for ws

and

have initial values of 1 and 0.11, respectively and

rest of the state’s initial value are set to 0. To combine

the incoming effect on any state, a number of options

available for choosing among the different

combination function. In the proposed model, the

identity and advanced logistic sum combination

alogistic

σ,τ

(…) functions are used as the standard

combination function. The parameters for

combination functions (

τ, σ )and speed of factors (η) of

the states for the scenario (discussed below) are given

in Table 3.

4 SCENARIO AND SIMULATION

RESULTS

Consider an office situation where an employee,

Mavrik works in front of a computer from 9 to 5. He

is not aware of the severity and susceptibility of being

sedentary and do not have the self-belief that he

should overcome this behaviour. The organization

started a campaign (intervention) for providing

information to the employees about the negative

consequences of sitting more than consecutives 30

minutes and offered them 10 minutes break after each

1 hour of work (reward). With this campaign, the

employee starts realizing the severity and

susceptibility of the prolonged sitting, which leads to

an increase in his attitude towards taking a break after

each hour. Now with the campaign, the employee

starts perceiving the benefit of walking and his

efficacy gets increased, so it reduces the sedentary

behaviour. To simulate the above scenario, the

connection values and the combination functions are

described in the table 2 & 3.

The parameter values shown in these tables can be

used to reproduce the results shown in fig. 2, 3 and 4

below. Moreover, only ws

and ws

have initial values

of 1 and 0.11 respectively. States with zero values for

τ and σ in table 3 suggest that Identity function has

been used for these sates.

Fig 2 displays all the states of the model. It can be

seen that initially the person’s negative belief about

his current sedentary behavior is almost zero. Which

means, the perceived susceptibility and severity of his

sedentary behaviour are also very low. Therefore, the

person keeps sitting for long time in office.

In the second half of the fig. 2, it’s observable that

a shift takes place in the dynamics of the states. This

shift is because of the intervention proposed for the

person for breaking the continuity of his sedentary

action. This intervention changes the person’s belief

by making him walk/move after certain amount of

time.

Computational Model for Changing Sedentary Behavior through Cognitive Beliefs and Introspective Body-feelings

447

Table 2: The connection values (



) between the two states.

Connection Weight Connection Weight Connection Weight





















0.8











0.5





















0.75











0.5











0.2





.





-0.2











0.8















.





-0.2











-0.6





















-0.8











0.8





















0.6









.

0.4











0.4



















.

0.45











0.9











-0.8









.

0.48











0.6





.





-0.6









.

0.55











-0.6





.





0.3











0.3















.





-0.6



















.

0.1





.





0.3











0.5









.

0.1





















0.8











0.3





















0.8

Table 3: The parameter of alogistic

σ,τ

(…) combination function and speed factor for different states.

State

τ σ η

State

τ σ η



0 0 0

srs

.

0.25 8 0.5



0 0 1



0.2 3 0.4

srs



0.4 8 0.2



0.3 10 0.4



0 0 0.5

srs

.

0.3 8 0.5



0 0 0.5

srs

.

0.3 8 0.5



0 0 0.5



0.2 8 0.02



0 0 0.5



0.2 8 0.02



0.35 8 0.4



0.3 8 0.2

srs



0.35 8 0.4

srs



0.3 8 0.1



0.35 8 0.4



0.3 7 0.2



0.4 8 0.4



0.28 7 0.1

esb



0.4 8 0.4



0.3 8 0.1

srs

.

0.25 8 0.5

Figure 2: The simulation result of the model with the state values mentioned in table 2 & 3.

0 25 50 75 100 125 150 175 200

Time

0.2

0.4

0.6

0.8

States Values

dynamic states

wss

sss

srss

psa1

esa1

psa2

esa2

ssb+

srsb+

fsb+

psb+

esb+

srsp.susptblty

srsp.svrty

bs-

bs+

srsp.efficacy

srsp.benefit

wsi

ssi

ssb-

srsb-

fsb-

psb-

esb-

HEALTHINF 2021 - 14th International Conference on Health Informatics

448

Figure 3: The sedentary behavior (esa(sedentary)) is high and the walking behavior (esa(active)) is low, but latter after with

intervention effect, the sedentary behavior went down and walking behavior got high.

Figure 4: All the cognitive belief states. The susceptibility and severity were initially too low and after the intervention it gets

high.

Fig. 3 shows that initially the person’s negative

belief (bs

) about his sedentary behaviour is low and

positive belief (bs

) is high. Hence his sedentary

action (ps

,es

) are high. In fig. 4, when the

intervention ws

and ss

gets activated to an enough

level, which changes the person’s belief, the

0 25 50 75 100 125 150 175 200

Time

0.2

0.4

0.6

0.8

States Values

dynamic states

wss

psa(sedentary)

esa(sedentary)

psa(active)

esa(active)

bs-

bs+

ssi

0 25 50 75 100 125 150 175 20

Time

0.2

0.4

0.6

0.8

States Values

dynamic states

wss

psa1

esa1

psa2

esa2

fsb+

srsp.susptblty

srsp.svrty

bs-

bs+

srsp.efficacy

srsp.benefit

wsi

ssi

fsb-

Computational Model for Changing Sedentary Behavior through Cognitive Beliefs and Introspective Body-feelings

449

perceived susceptibility srs

p.susptblty,

and severity

srs

p.svrty

of the sedentary action get increased.

Negative belief (bs-) increases and positive belief

(bs+) decreases. As a result, the person’s non-

sedentary action i.e., preparation for action ‘2’ ps

and execution of action ‘2’ es

also increases while

sedentary behaviour i.e., preparation for action ‘1’

and execution of action ‘1’ es

decreases

REFERENCES

Ajzen, I. (2005). Attitudes, personality, and behavior.

McGraw-Hill Education (UK).

Australian Governement guidleines for sedenteray

behavior. (2019). Retrieved from http://www.

health.gov.au/internet/main/publishing.nsf/Content/sbe

haviour

Bandura, A. (1998). Health promotion from the perspective

of social cognitive theory. Psychology and Health,

13(4), 623–649.

Bandura, A. (2004). Health promotion by social cognitive

means. Health Education & Behavior, 31(2), 143–164.

Damasio, A. R. (1998). Emotion in the perspective of an

integrated nervous system1. Brain Research Reviews,

26(2–3), 83–86.

Direito, A., Carraça, E., Rawstorn, J., Whittaker, R., &

Maddison, R. (2016). mHealth technologies to

influence physical activity and sedentary behaviors:

behavior change techniques, systematic review and

meta-analysis of randomized controlled trials. Annals

of Behavioral Medicine, 51(2), 226–239.

Dunn, E. E. (2017). Sedentary Behaviour, Physical

Activity, and Mobile Apps Among University Students.

Hadgraft, N. T., Winkler, E. A. H., Healy, G. N., Lynch, B.

M., Neuhaus, M., Eakin, E. G., Fjeldsoe, B. S. (2017).

Intervening to reduce workplace sitting: mediating role

of social-cognitive constructs during a cluster

randomised controlled trial. International Journal of

Behavioral Nutrition and Physical Activity, 14(1), 27.

Huang, Y., Benford, S., & Blake, H. (2019). Digital

Interventions to Reduce Sedentary Behaviors of Office

Workers: Scoping Review. Journal of Medical Internet

Research, 21(2), e11079. https://doi.org/10.2196/11079

Michie, S., Richardson, M., Johnston, M., Abraham, C.,

Francis, J., Hardeman, W., Wood, C. E. (2013). The

behavior change technique taxonomy (v1) of 93

hierarchically clustered techniques: Building an

international consensus for the reporting of behavior

change interventions. Annals of Behavioral Medicine,

46(1), 81–95. https://doi.org/10.1007/s12160-013-

9486-6

Owen, N., Sugiyama, T., Eakin, E. E., Gardiner, P. A.,

Tremblay, M. S., & Sallis, J. F. (2011). Adults’

Sedentary Behavior. American Journal of Preventive

Medicine, 41(2), 189–196. https://doi.org/10.1016/

j.amepre.2011.05.013

Prapavessis, H., Gaston, A., & DeJesus, S. (2015). The

Theory of Planned Behavior as a model for

understanding sedentary behavior. Psychology of Sport

and Exercise, 19, 23–32.

Sallis, J. F., Owen, N., & Fisher, E. (2015). Ecological

models of health behavior. Health Behavior: Theory,

Research, and Practice, 5, 43–64.

Scherer, K. R. (2005). What are emotions? and how can

they be measured? Social Science Information, 44(4),

695–729. https://doi.org/10.1177/0539018405058216

Taj, F., Klein, M. C. A., & van Halteren, A. (2019). Digital

Health Behavior Change Technology: Bibliometric and

Scoping Review of Two Decades of Research. JMIR

MHealth and UHealth, 7(12), e13311.

Taj, F., Klien, M., & Aart Van Halteren. (2020). An Agent-

Based Framework for Persuasive Health Behavior

Change Intervention. In 9th International Conference

on Health Information Systems (HIS 2020) 20-23

October 2020 (pp. 157–168). Amsterdam and Leiden,

Netherlands: Springer Lecture Notes in Computer

Science (LNCS). https://doi.org/10.1007/978-3-030-

61951-0_15

Treur, J. (2016). Dynamic modeling based on a temporal-

causal network modeling approach. Biologically

Inspired Cognitive Architectures, 16, 131–168.

https://doi.org/10.1016/j.bica.2016.02.002

Weggemans, R. M., Backx, F. J. G., Borghouts, L.,

Chinapaw, M., Hopman, M. T. E., Koster, A.,

Committee Dutch Physical Activity Guidelines 2017.

(2018). The 2017 Dutch Physical Activity Guidelines.

International Journal of Behavioral Nutrition and

Physical Activity, 15(1), 58. https://doi.org/10.1186/

s12966-018-0661-9.

HEALTHINF 2021 - 14th International Conference on Health Informatics

450