881

1 INTRODUCTION

The Baltic Sea is an important socio-economic zone

with countries in the Baltic Sea Region (BSR)

depending on it for movement of traffic and goods all

year round. The northern part of the Baltic Sea, the

Bay of Bothnia, experiences strong winters resulting

in ice-covered waters for nearly 5 months each year

[1]. To ensure the safety and efficiency of traffic in

winter in this region, icebreakers are often employed

to assist vessels. The icebreakers are a critical and

shared resource, often jointly managed by the traffic

authorities of neighbouring countries (such as Finland

and Sweden). The icebreakers perform many tasks

during the months from October to May, when the sea

is usually covered in ice. Icebreakers create and

maintain channels in ice called directed pathways

(dirways) which are then used by other vessels, that

are not designed for icebreaking capabilities, to

operate in. Icebreakers also tow vessels with lesser

icebreaking capabilities through regions with tougher

ice conditions. They help vessels navigate the tricky

fast ice region closer to the ports [2]. Operating,

maintaining, and coordinating icebreakers are

expensive [3]. Hence, the icebreakers always try to

optimize their trips, assisting multiple vessels at a

time whenever possible and reducing their own travel

time. It is often the case that the number of vessels

requiring assistance at a given moment are more than

the number of icebreakers in service. The icebreakers

are then tasked with prioritizing the assistance

requests. These priorities need to consider multiple

factors such as average waiting time, fuel

consumption, distance to be travelled, expected

departure/arrival time of vessels from/to ports, and

safety requirements. Icebreaker decision-making is

Cognitive Task Analysis to Understand Icebreaker

ecision Making

M. Musharraf

, K. Kulkarni

& S. Mallam

2,3

Marine and Arctic Technology, Aalto, Aalto University

Memorial University of Newfoundland, St. John's, Canada

University of South-Eastern Norway, Borre, Norway

ABSTRACT: Icebreakers are a critical shared resource between Finland and Sweden required to keep the Baltic

Sea clear for differing waterborne activities throughout the winter season. The safety and efficiency (both

ecological and in terms of time) of the winter transport system is highly dependent on the decision-making

process followed by the icebreakers. The captains in charge of the icebreakers must decide the priority of

assistance among all the merchant vessels that may be in need within a given time window. While captains

successfully do this every year, with experience, this decision-making becomes second nature and a transparent

picture of how the decisions are made is often missing. It is not always clear what salient features captains pay

attention to, and how they use those features in their reasoning process to reach a decision during operations.

This paper presents a pilot study that uses cognitive task analysis (CTA) to outline captains' decision-making

process for ice breaker assistance allocation. In-depth interviews of three subject matter experts were conducted

using a naturalistic icebreaker scenario. Results include identified critical decision points, identified, and

prioritized salient features, and characterized icebreaker assistance strategies.

http://www.transnav.eu

the

International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 17

Number 4

December 2023

DOI: 10.12716/1001.17.04.

882

thus a complex task, involving multiple dynamic

factors and system-level Key Performance Indices

(KPIs) [3]. These decisions are currently managed by

the icebreaker captains with the support of other

expert seafarers on land [4].

In the recent years, maritime traffic authorities are

additionally faced with challenges emerging from

climate change [1]. On the one hand, the winters are

more unpredictable with more brash and slushy ice,

and on the other hand, vessels are now required to be

Energy Efficiency Design Index (EEDI) compliant for

fuel efficiency, which results in lesser ice breaking

capabilities. The policymakers are now interested to

understand how these changes are set to affect the

icebreaking needs in the coming future. This has

resulted in multiple research efforts, especially in the

BSR universities, to study and predict the future of

icebreaking under climate change challenges [5][6].

While many of these works have been successful at

capturing the ice dynamics, the traffic flows, and the

vessel performances, the one aspect that has been

hard to model is the icebreaker decision-making.

Previous attempts at capturing this decision-

making process have resulted in logic-driven and/or

mathematical algorithms [6]. However, when

compared with historical data during validation

studies, large gaps were identified in the actual data

and the modelled versions. One reason for this gap is

that the purely logic-driven approaches fail to count

the human factors (such as the on-the-job adjustments

made by experts to suit dynamic situations) that are

involved in the decision making. This paper uses a

cognitive task analysis (CTA) approach to elicit

knowledge about icebreaker decision-making from

the subject matter experts. This allows capturing the

human element in the decision making.

A pilot study with three participants has been

conducted. The participants were recruited based on

their experience with the Finnish-Swedish winter

navigation system. Among the available CTA

methods, the Critical Decision Method (CDM) [7] has

been used with a predesigned naturalistic ice-breaker

scenario and predefined probes. While the results of

the study are preliminary, they show that the CDM

method can be useful in identifying critical decision

points, identifying, and prioritizing salient features,

and characterizing the strategies used. The outcome of

this study is expected to increase the realism of winter

navigation simulation tools that involve icebreakers,

contribute to intelligent decision support systems for

winter navigation, and generate training materials

that will be useful for less experienced seafarers

tasked with ice breaker decision making.

2 COGNITIVE TASK ANALYSIS AND THE

CRITICAL DECISION METHOD

CTA is an approach implemented to uncover what

people know and how they think, to better

understand the mental processes that underlie

observable behaviour [8]. CTA describes a collection

of differing methods that are an extension of the

traditional task analysis approach to describe

knowledge, thought-processes, mental strategies, and

goal structures of actors in complex systems [9].

Among the many CTA methods available, CDM is

used in this paper. The CDM aims to achieve

knowledge elicitation through cognitive probing and

reflection as a form of retrospective CTA technique

[10]. CDM is commonly used to elicit specialized

knowledge from subject matter experts across a

diverse number of domains to better understand

expert decision making and reasoning in naturalistic

settings [11]. Although CDM was primarily intended

for non-routine events [10] CDM can also be applied

to both routine and non-routine analysis of highly

specialized or difficult events and tasks, especially

when decision-making and actions of experts may

differ from those with less experience [12].

CDM is implemented through a semi-structured

interview approach, typically consisting of seven

steps: i) define the task or scenario under analysis, ii)

select CDM probes, iii) select appropriate participant,

iv) gather and record account of the incident, v)

construct incident timeline, vi) define scenario phases

or decision points, vii) use CDM probes to query

participant decision making. This study uses the

typical CDM steps with some modifications. Steps iv

and v which are purposefully designed to analyse

non-routine retrospective incidents, were eliminated

from the current study as the focus is on typical ice-

breaking scenarios. The other steps also needed some

small adjustments to fit the purpose of the study. The

steps are described in the following subsections with

more details.

2.1 Define the task or scenario under analysis

The first step focuses on defining the task or scenario

under analysis. Unlike typical CDM applications

where focus in on non-routine incidents, the current

study focuses on a typical ice breaking scenario.

However, ice breaking itself is a highly specialized

task and even for typical scenarios the decision-

making is quite complex in nature.

The scenario used in the study focuses on the

Northern Baltic Sea, the Bay of Bothnia region. The

scenario is a snapshot of a situation in winter, in line

with some of the commonly occurring instances in

winter navigation. During scenario development,

academic collaborators with prior experience of

working with seafarers were consulted for

understanding what information are crucial to design

a credible scenario. Figure 1 shows the scenario that

was presented to the participants. Four vessels are

shown to exist in the system (that is, the area under

observation: Bay of Bothnia) at the time of

observation. The vessels are assigned details such as

ice class and name. Discussion with academic

collaborators during the scenario design phase

revealed these details as crucial. While other ship

details (such as hull type and propulsion power) were

also deemed important it was concluded that captains

and seafarers who work in the region on a regular

basis can infer the other vessel details from the vessel

name. The vessel name and ice class were assigned

from a set of vessels that frequently navigate this area

using public data sources [13]. The locations and

departure times of the vessels were inspired from

instances that occurred during simulation runs of a

winter navigation simulation tool [6] developed at

Aalto University. The tool uses inputs from

Automatic Identification System (AIS) data for the

year 2018. Additional information that was

883

Figure 1. a) Scenario under analysis with 4 vessels and 2 icebreakers b) Symbols for ice information in (a).

identified as crucial was the prevailing ice condition.

Historical ice data [14] from the Finnish

Meteorological Institute (FMI) were used to represent

the ice conditions. In practice and in the simulation,

the icebreakers operate in their assigned zones,

prioritizing the vessels in their zone. This information

is not drawn explicitly in any chart and the zones are

decided dynamically by icebreaker captains.

In Figure 1(a), V indicates vessel and IB indicates

icebreaker. Both IBs have the same capacity to create a

wide enough channel. The black directed arrows at

the vessels’ end indicate the direction of intended

travel of the vessel. The pink lines indicate channels.

Single line indicates closed channels. Two lines

indicate open channel. Some channels are partially

open (near V1), some are fully open (near V2).

Narrowing width of channels indicates that the

channel is progressively closed. Figure 1(b) shows

the interpretation of the symbols used for ice

information in 1(a).

Given the scenario, the participants were tasked to

prioritize the vessels for assistance assuming they

were in charge of the icebreakers. With assistance

from the analysts, the participants described how they

would use the icebreakers to help the vessel in need

until all vessels are either safely at a destination port

or are safely navigating.

2.2 Select CDM probes

Since the aim of the study was to identify and

prioritize salient features and characterize the

icebreaker assistance strategies, the CDM probes were

designed accordingly. The set of probes relevant for

this paper is presented in Table 1.

Table 1. CDM probes for investigating ice-breaker decision

making

________________________________________________

1. Goal What were your specific goals at this

specification decision point?

2. Cue What features were you looking for when

Identification you formulated your decision?

3. Information What was the most important piece of

integration information that you used to formulate the

decision?

4. Situation Except for the information given to you,

assessment was there any additional information that

you might have used to assist in the

formulation of the decision?

5. Basis of Do you think that you could develop a

choice rule, based on your experience, which

could assist another person to make the

same decision successfully?

Why/Why not?

________________________________________________

2.3 Select appropriate participant

Since this is a pilot study, the participant pool was

limited to 3. The participants were recruited based on

their experience with the Finnish-Swedish winter

navigation system. Inclusion criteria required that

participants were certified nautical officers and had

experience with ice navigation in the Baltic Sea. The

participants had between 17-25 years’ experience

working at sea in various operational positions, with

two of the participants also having additional

management experience in planning and executing

winter navigation. Define decision points and use of

CDM probes to query participant decision making

The decision point identification was done in

conjunction with the participants. It was agreed that

every time the icebreaker is taking a new decision

(e.g., deciding on a new vessel to assist, deciding to

form a convoy) it can be called a new decision point.

However, all participants agreed that their answers to

probe questions would not change for the different

884

decision points. Keeping this in mind, the probes

presented in Table 1 were used once per participant

per scenario rather than per decision point.

3 DATA ANALYSIS AND RESULTS

To increase the reliability of the interview process and

data collection, two analysts were used throughout

the CDM steps mentioned in section 2. Several data

were collected but the data that has been analysed for

the purpose of this paper includes notes from the

analysts and notes (along with drawings) from the

participants (if any were produced during the CDM

interview process). While audio recordings were

taken for each participant, they have only been used

in this paper to resolve conflicts that arose while

comparing notes of the two analysts, as a form quality

assurance and inter-rater reliability. Two direct

outcomes of the analysis were 1) the participants’

answer to the question regarding prioritization of

vessel as described in section 2.1 and 2) a CDM table

with the participants’ answers to the questions listed

in Table 1. The CDM table can be further analysed to

identify and prioritize salient features, and

characterize the strategies used. One indirect outcome

was an evaluation of the credibility of the scenario

and suggestions on how the scenario can be further

improved for a future full-scale study.

The following subsection summarizes the most

significant outcomes of the study.

3.1 Prioritizing vessels for assistance

For the scenario presented in Figure 1, participants

discussed how they would prioritize the vessels in

need and plan assistance until all vessels are either

safely in port or are safely navigating.

For the first assistance decision (decision point 1),

some similarities across participants were observed.

All participants agreed that IB2 should assist V4 first.

Regarding IB1, two participants chose that IB1 should

assist V3 first. Both participants mentioned that V1

and V2 can sail independently at least for a while

given their ice class. The other participant used IB1 to

assist V2 first. At the next decision point, again some

similarities in decision making were observed across

participants. Two participants mentioned that V3 and

V4 may construct a convoy that can be assisted by IB2

until the zone crossing and then the convoy can be

handed over to IB1. One of these participants

mentioned that V2 can also join this convoy later.

While V1 was identified to not need help, two

participants mentioned that it can be assisted by IB1 if

needed but only after the other assistance has been

taken care of (so not a priority), and at the zone

crossing IB2 can take over V1.

3.2 Salient features and strategies used

Questions 1 and 5 in Table 1 focused on the strategies

of icebreaker assistance while Questions 2 to 4 focused

more on identifying and prioritizing salient features.

When asked about goal specification (Question 1,

Table 1), it was realized that traffic safety is of utmost

importance and the goal is always to ensure this.

Given that this is usually already considered in terms

of ice class of the ships and assistance restriction, the

goal in hand is to minimize the overall waiting time of

e vessels. Next goal would be to optimize fuel

consumption, and this goes hand in hand with

optimizing overall waiting time of vessels. All

participants had the same view on this question. The

goals specified by the participants and their priorities

are presented in Table 2.

Table 2. Goal specifications and priorities

________________________________________________

Goal Priority

________________________________________________

Traffic safety 1

Minimize overall waiting time for vessels 2

Optimize fuel consumption 2

________________________________________________

Regarding the features that are important for

decision formulation (Question 2, Table 1),

participants had slightly different views. All

participants identified Expected Time of Departure

(ETD) of the vessels as an important feature. Ice

condition and ice class of vessels were identified as

important features by two participants. It is worth

noting that while the other participant did not directly

mention ice condition and ice class as important

features, they did use them while formulating the

basis of choice in Question 5. Zone of icebreakers, size

of the vessels (i.e., are these technically good vessels),

dirway location, and location of icebreakers with

respect to vessel were other features that were singly

mentioned by the participants.

Arguably, the most important piece of information

needed to formulate the decision (Question 3, Table

1), two participants mentioned ETD of the vessels.

The other participant identified ice condition as the

most important information.

The question regarding gaps in the provided

information (Question 4, Table 1) was appreciated by

the participants. The most critical information that

was identified to be missing was the wind

information. The wind direction, strength, and

pressure dictate the dynamic ice condition (such as ice

drift). Two participants mentioned that based on the

wind information one may almost have a new ice

chart. In practice, with access to Icebreaker Net

(IBNet), the icebreaker captains can create this new ice

chart. The wind information along with the

temperature also dictates how quickly or slowly a

new channel will close. Port information (such as port

calls and berthing availability) was missing, and it

was also identified as a critical information.

While the use of ice chart was deemed okay, it was

mentioned by the participants that in real life the

captains will have access to satellite images through

IBNet. In the absence of IBNet, it was suggested that

instead of using a static ice chart, several ice charts

(perhaps from a few days before) should be provided.

The icebreaker captains also have access to fairway

traffic information including how frequently the

fairway been visited in recent time. This helps to

identify if the ice in the fairway is new, recently

broken, or has hardened over time. The nature of

assistance is affected by this information. Other

missing information includes assistance restriction on

885

certain ice classes for certain ports, history of a ship

for this winter: is it a frequent caller (already have a

history on how they have been assisted this winter),

ice field updates and other records from pilots put in

the IBNet, records of phone calls to pilots, more

specific vessel details such as propulsion power and

hull type. The captains also have access to past tracks

(knowledge of icefields, where channels have been

created), and in-field information about the location of

difficult spots in ice.

The last question was about the basis of choice

where participants were asked to develop a rule to

assist another person to make the same decision

successfully (Question 5, Table 1). There is some

similarity in the rules that participants 1 and 2

developed. The first step for both participants was to

identify the vessels in each icebreaker zone that may

require assistance. Vessel ice class and prevailing ice

condition was used to assess the assistance

requirement. Participant 1 also mentioned that for

vessels in port, we know how they have been able to

get there so chances are that they will be able to get

out in similar manner with similar assistance

requirements (depending on fairway situation). Then

within each IB zone, participant 1 proposed using a

first come first serve rule. This participant mentioned

including incoming traffic while doing this.

Participant 2 proposed choosing a vessel in need that

is closest to the icebreaker. This participant added that

the ETD of the assisted vessel dictates icebreaker’s

own departure time. Icebreaker will adjust its own

engine power based on the departure time. If there is

time, there is no need to go full throttle. Icebreakers

will meet the vessels as close as possible to the point

at which it is likely to be needing assistance. The goal

is to minimize icebreaker movement.

Participant 3 had a bigger picture perspective

based on choice. For non-critical situations, the

participant had a rule to focus on minimizing the

waiting time. The participant pointed out “[As IB

captain] You have zones, you know each other, you

know vessels in your zone, and you know the

incoming traffic.” Based on this, the waiting time

minimization can be done. For safety critical scenarios

like having high ice pressure, the participant advised

to prioritize safety over minimizing waiting time. As

mentioned by the participant “It takes guts to make

safe decision since you may be increasing cost and

waiting time.”

Table 3. Basis of choice: rules provided by participants

________________________________________________

Parti- Rule

cipant

________________________________________________

1 First use vessel ice class and prevailing ice

condition to decide who needs assistance. For each

IB zone, follow first come first serve. Look at

incoming traffic in the zone.

2 The ice condition (at departure time) and vessel ice

class suggest where vessel might get stuck. Then

the IB chooses the one that that is closest to its own

location. Vessel ETD dictates IB’s own departure

time. Adjust engine power based on dept time.

Meet the vessel as close as possible to the point at

which it is likely to be stuck.

3 If you don’t have any pressure, focus on

minimizing the waiting time. If there is pressure,

prioritize safety. It takes guts to make safe decision

since you may be increasing cost and waiting time.

________________________________________________

The basis of choice and the rules provided by each

participant are presented in Table 3.

3.3 Credibility of the scenarios and suggestions for

improvement

Question 4 reveals the information that are important

for assistance decision making but were missing from

the scenario description. The participants suggested

that discussion on this should be continued as it is

possible to include at least some of the missing

information quite easily. They offered to help in

increasing the credibility of the scenario. Besides the

identified missing information, all participants

mentioned adding a 3rd icebreaker to the South to

make the scenario more realistic. One participant

mentioned the missing wind information as critical

and had to assume a wind direction to proceed with

the scenario. This participant also evaluated the

dirways to be unrealistic. This is because although the

current location of dirways is set in easier ice

conditions, it is too close to the shore. In practice,

dirways are located further away from the shore, even

if it means breaking channels in harder ice. This is to

prevent grounding incidents. The participant

suggested to refer to IBNet data and consult with

experts in drawing dirways for future scenarios.

4 POTENTIAL APPLICATION OF THE RESULTS

Icebreaker decision-making is an important and

elusive piece of the winter navigation operations that

greatly affects its efficacy. The outcome of this study is

expected to increase the realism of the ice-breaker

behaviour in the simulation tool developed by Aalto

University in close co-operation with the Finnish

Transport Infrastructure Agency. This will enable

realistic evaluation of several “what-if” scenarios,

including engine power and ice-breaker scheduling

optimization for safe, efficient, and environmentally

friendly winter navigation. The outcome of the study

will also contribute to developing intelligent systems

that will support decision making for winter

navigation. Given the nature of this study and the

necessary detailed documentation, the results are also

expected to generate educational materials, which can

be used for training less experienced decision-makers

and seafarers. The results are also expected to bring

transparency to a process that has otherwise been

hard to understand for those not directly involved in

it. This could lead to better trust overall in the

navigation environment and facilitate healthier

cooperation between all stakeholders.

5 CONCLUSIONS AND FUTURE WORK

This paper describes a novel attempt at using CDM to

decipher the complex icebreaker decision-making

process. A pilot study was conducted to test the

efficacy of this approach in this problem domain. The

study involved subject matter experts in winter

navigation. The results brought forth multiple

interesting facets of the decision-making process that

886

were hitherto unidentified by prior research efforts.

The participants also gave several important inputs on

how the scenarios can be improved for more directed

knowledge elicitation. The future work will involve

conducting a full-scale study with a larger number of

participants with inter-rater reliability analysis and

with refined operational scenarios that are more

comprehensive and more realistic.

ACKNOWLEDGEMENT

This work was supported by the Academy of Finland

project: Towards human-centered intelligent ships for

winter navigation (Decision number: 351491)

REFERENCES

[1] https://en.ilmatieteenlaitos.fi, “Ice conditions,” Finnish

Meteorological Institute. https://en.ilmatieteenlaitos.fi

(accessed Mar. 31, 2023).

[2] Rosenbland, M., 2007. Increasing the Safety of Icebound

Shipping, Final Scientific Report: Volume 1, Helsinki

Uni-versity of Technology, Ship Laboratory, Espoo,

Finland. WP4. Operative Environment (Icebreaker

Operations).

[3] Baltic Icebreaking Management, 2021. Baltic Sea

Icebreaking Report 2020-2021.

https://baltice.org/app/static/pdf/BIM%20Report%20202

0-2021.pdf.

[4] Finnish Transport Infrastructure Agency, 2021. Finland’s

winter navigation. Instructions for winter navigation

operators. Helsinki.

[5] Kondratenko, A. et. al. Decarbonizing shipping in ice by

intelligent icebreaking assistance: a case study of the

Finnish-Swedish winter navigation system (manuscript

ready for submission)

[6] K. Kulkarni, P. Kujala, M. Musharraf, and I. Rainio,

“Simulation Tool for Winter Navigation Decision

Support in the Baltic Sea,” Applied Sciences, vol. 12, no.

15, p. 7568, Jul. 2022, doi: 10.3390/app12157568.

[7] Crandall, B., Klein, G. A., & Hoffman, R. R. (2006).

Working minds: A practitioner's guide to cognitive task

analysis. Mit Press.

[8] Schraagen, J. M., Chipman, S. F. & Shalin, V. L. (2000).

Cognitive Task Analysis. Lawrence Erlbaum Associates,

Inc.

[9] Stanton, N., Salmon, P. M., & Rafferty, L. A. (2013).

Human factors methods: a practical guide for

engineering and design. Ashgate Publishing, Ltd.

[10] Klein, G. A., Calderwood, R. & Macgregor, D. (1989).

Critical decision method for eliciting knowledge. IEEE

Transactions on systems, man, and cybernetics, 19(3),

462-472.

[11] Klein, G., & Armstrong, A. A. (2004). Critical decision

method. In Handbook of human factors and ergonomics

methods (pp. 373-382). CRC Press.

[12] Hoffman, R. R., Crandall, B. & Shadbolt, N. (1998). Use

of the critical decision method to elicit expert

knowledge: A case study in the methodology of

cognitive task analysis. Human factors, 40(2), 254-276.

[13] https://www.marinevesseltraffic.com/2013/06/mmsi-

number-search.html (accessed Mar. 31, 2023).

[14] Ice data: https://www.smhi.se/oceanografi/istjanst/

havsis_en.php (accessed Mar. 31, 2023).