Characterizing Machine Guidance in Geospatial Analysis

Yue Hao

and Guoray Cai

College of Information Sciences and Technology, The Pennsylvania State University, University Park, PA, U.S.A.

{yqh5470, gxc26}@psu.edu

Keywords:

Geographic Information Systems, Geospatial Analysis, Machine Guidance.

Abstract:

Geospatial analysis poses challenges for individuals with limited expertise in Geographic Information Science

(GIScience) methods and tools, requiring complex decision-making and spatial reasoning, often leading to

difﬁculties or even failures. This research explores machine guidance as a novel approach to support domain

analysts throughout the geo-analytical process. This approach can be implemented as an intelligent inter-

face agent that is capable of recognizing user’s analytical difﬁculties and providing proactive assistance. We

propose a conceptual framework that characterizes the cooperative role of machine guidance in geospatial

analysis. We focus on answering two key research questions: (1) when to guide? and (2) how to guide?. The

framework provides a foundation for future research on machine-guided geospatial analysis, informing the de-

velopment of other computer-aided systems that enhance usability and analytical effectiveness in GIScience.

1 INTRODUCTION

Geospatial analysis is crucial to scientiﬁc investiga-

tion and decision-making in various domains (Good-

child and Longley, 1999), ranging from human-

environment interactions (De Smith et al., 2007), un-

derstanding the health impact of COVID-19 and vac-

cines (Sun et al., 2020; Wang et al., 2021; Mol-

lalo et al., 2021), healthcare resources (Kang et al.,

2020), to public health policies (Ram

ırez and Lee,

2020; Ahasan and Hossain, 2021). These demands

for geospatial analysis stimulated the rapid growth

of analytical tools and methods (Goodchild et al.,

2000). As Geographic Information Systems (GISys-

tems) grow into a type of high-functionality system, it

becomes increasingly harder to learn and use (Lieber-

man et al., 2015).

Despite its signiﬁcance, geospatial analysis

presents challenges due to its complexity. It involves

processing and analyzing geographic data by repre-

senting spatial phenomena, utilizing tools and statis-

tical methods, and interpreting spatial relationships

(Bailey et al., 1995; Goodchild, 2006). Analysts must

recognize patterns in spatial data and understand their

underlying signiﬁcance (Haining, 1994). At the tech-

nical level, geospatial analysis makes use of a variety

of analytical tools and techniques to understand geo-

graphic patterns and events (Goodchild, 1992). How-

ever, these tasks are often cognitively demanding and

https://orcid.org/0000-0002-8368-4924

require specialized knowledge and skills in geography

and GISystems, which are not universally accessible.

The complexity of geospatial analysis tools, com-

bined with the need for GIS expertise, creates barriers

for analysts, particularly those outside the GIS ﬁeld.

Existing solutions primarily focus on two approaches.

The ﬁrst approach involves empowering analysts by

offering education and training programs in geogra-

phy and GIScience (Council et al., 2005). While these

programs are valuable, they are not scalable due to

their high costs in terms of time and money. Sec-

ond, GISystem user interfaces have been enhanced

to better reﬂect how humans perceive, interact with,

and conceptualize the world to improve usability and

analytical capabilities (Goodchild, 2009). However,

even with improved interfaces, analysts still face sig-

niﬁcant burdens, such as forming effective analytical

strategies, preparing data, and executing the required

spatial functions. These limitations highlight the need

for innovative solutions to make geospatial analysis

more accessible and efﬁcient. To make geospatial

analysis practical to a diverse range of applications

and analysts, solutions are needed to bridge the cog-

nitive and skill gap.

Motivated by addressing the human cognition and

expertise barriers during spatial analysis, we propose

machine guidance (Ceneda et al., 2018) as a novel

approach to make geospatial analysis more accessi-

ble. The key idea of our approach is to introduce an

intelligent interface agent that offers timely help and

Hao, Y. and Cai, G.

Characterizing Machine Guidance in Geospatial Analysis.

DOI: 10.5220/0013240300003935

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 11th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2025), pages 39-50

ISBN: 978-989-758-741-2; ISSN: 2184-500X

guidance when analysts encounter difﬁculties in ad-

vancing their analytical goals. A machine guidance

agent is able to recognize when users are facing dif-

ﬁculties and take the initiative to guide them with

appropriate actions (Liu et al., 2018; Ceneda et al.,

2019). By leveraging the guidance agent’s expertise,

human analysts’ efforts and cognitive load in identify-

ing clues or suitable solutions when facing challenges

can be reduced. Thus, the machine guidance agent is

particularly valuable for domain experts who are pro-

ﬁcient in their speciﬁc problem domain but with lim-

ited expertise in GISystems or GIScience (Nyerges,

1995; Traynor, 1998).

Effective machine guidance agent requires two

key capabilities. First, the guidance agent must be

able to recognize when analysts are facing challenges

and whether guidance is needed. Second, it must be

able to determine what types of guidance will be pro-

vided based on the nature of user challenges. The

overall goal is to make an analytic process produc-

tive and effective even in cases where users’ exper-

tise is inadequate. Towards deﬁning the conceptual

framework for machine guidance in geospatial analy-

sis, we explore two key questions to inform its design

and implementation:

1. When is guidance needed during the geospatial

analytical process? This involves understanding

the nature of geospatial analysis process and iden-

tifying the points when users might face chal-

lenges.

2. How should guidance be provided? Answering

this question involves understanding alternative

messages and media for communicating guidance

and how to choose them to ensure effectiveness

and keep them minimally intrusive.

By addressing these two research questions, we

aim to characterize a conceptual framework for ma-

chine guidance in geospatial analysis. We begin by

conceptualizing the geoanalytical process to identify

when guidance is most needed. Using a hypotheti-

cal scenario, we illustrate how machine guidance can

be integrated into the analytical process in various

forms. A better understanding of when and how guid-

ance should be designed contributes to the develop-

ment of a science of design for machine guidance.

It can ﬁll the gap that there is a lack of understand-

ing of the machine guidance behaviors in facilitating

geospatial analysis. It also can be reused or further

developed regarding the mixed-initiative guidance de-

GIScience agenda on the design and use of effective

machine guidance to amplify human capacity for spa-

tial problem-solving.

2 LITERATURE REVIEW

2.1 Conceptualize Geospatial Analysis

Geospatial analysis is a complex problem-solving

process that integrates spatial thinking to address sci-

entiﬁc questions and support decision-making (Good-

child and Longley, 1999; Fischer et al., 2011). Unlike

other analytical domains, geospatial analysis requires

viewing problems with a “geographical eye”, empha-

sizing spatial relationships, patterns, and interconnec-

tions (Downs, 1997). This process involves systemat-

ically handling spatial data and extracting meaningful

insights, ensuring that spatial objects and their rela-

tionships are accurately represented and understood

(Haining, 1994). As an iterative analytical method,

it involves multiple stages, including problem identi-

ﬁcation, strategy formation, analysis, and evaluation

(De Smith et al., 2007; Bednarz et al., 2013). We cat-

egorize this process into Figure 1. Each step demands

cognitive and computational efforts to effectively in-

terpret spatial phenomena, manage uncertainties, and

optimize solutions (Goodchild and Janelle, 2010).

The problem-solving aspect of geospatial analy-

sis requires structuring ill-deﬁned spatial problems

into well-represented phenomena using spatial con-

cepts such as distance, orientation, connectivity, and

scale (Huisman et al., 2009; Miller and Wentz, 2003).

Analysts must carefully choose external representa-

tions and operations based on the inherent uncer-

tainties in geographic data, as spatial representations

and scales inﬂuence analytical outcomes (Couclelis,

2003; Grekousis, 2020). Effective geospatial analy-

sis also depends on the strategic integration of data,

logical task sequencing, and cognitive reasoning to

synthesize and evaluate spatial information dynam-

ically (D

orner and Kreuzig, 1983; Thomas, 2005).

Since spatial problems lack predeﬁned solutions, iter-

ative reﬁnement, and expert judgment are necessary

to improve decision-making and ensure analytical ac-

curacy (Pretz et al., 2003; Bednarz, 2004).

Figure 1 illustrates the workﬂow and components

involved in geospatial analysis conducted by a human

analyst. Analysts, as domain experts, typically start

by deﬁning a problem within their speciﬁc ﬁeld. The

problem should then be translated into a spatial repre-

sentation based on the analytical phenomena and their

interactions. Next, the spatial problem can be bro-

ken down into a series of questions that GIS tools

can address. Addressing these questions often re-

quires executions and operations in a coherent work-

ﬂow within GISystems. Based on the GIS outputs,

the analyst compares and selects strategies. The ana-

lyst also needs to synthesize and evaluate the outputs

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management

Figure 1: The geoanalytical process that requires a series of tasks and decisions.

and intermediate results to actively evaluate and ad-

just their analytical workﬂow as well as the problem-

solving strategies.

2.2 Cognitive Challenges of Geospatial

Analysis and the Need for Support

Geospatial analysis includes an essential goal of mak-

ing critical decisions based on locations. It improves

one’s understanding of the geographic phenomenon

presented by the data by discovering and revealing the

previously unknown patterns (Huisman et al., 2009).

Cognitively, it is challenging to see things and situ-

ations spatially for structuring problems, ﬁnding an-

swers, and expressing solutions (Council et al., 2005).

Generating new knowledge from the result requires

human reasoning and understanding (Ward et al.,

1999). It requires thinking and operational skills from

three different dimensions: spatial concepts, spatial

representation, and spatial reasoning (Council et al.,

2005).

However, constraints of human perception and

cognition, such as limited ﬁeld of view in percep-

tion and working memory in cognition, have been

documented in psychological studies (Baars, 1997;

Baddeley, 1992; Creem-Regehr et al., 2005). These

limitations impact how we perceive and comprehend

complex data and situations. Choosing the suitable

representation, for instance, is a demanding task that

requires one’s cognitive efforts in deciding the ex-

plicit relation and structures of the space as well as

the unknown parts (e.g. data scale and aggregation

level) that one is going to explore (Freksa et al., 2017;

De Smith et al., 2007). Spatial representation should

match the spatial phenomenon as well (Dungan et al.,

2002). At an operational level, proper choice and

use of analytical tools are critical (De Smith et al.,

2007). Generated outputs entail spatial patterns that

vary greatly depending on particular methods that are

applied (De Smith et al., 2007). Making these critical

decisions is inherently a demanding process that re-

quires adequate training to gain spatial thinking skills

(Dramowicz et al., 1993). External assistance is of-

ten required in facilitating the analysts to go through

these procedures.

2.3 Inadequate Support to Geospatial

Analysis

Efforts to support an effective spatial analytical pro-

cess exist in two tracks. One track is to train ana-

lysts to think spatially about geographic phenomena

and their representations through education and train-

ing (Council et al., 2005). The second track is to im-

prove analytical tools and their interfaces in support-

ing geospatial analytical tasks.

GIScience education and geographical education

are interconnected and complement each other in pro-

viding students with a comprehensive understanding

of geography and the tools needed to address geo-

graphic challenges in various domains (Roche, 2014;

Shin et al., 2016). Based on the geographic foun-

dations, the concept of spatial thinking serves as a

framework for structuring problems, ﬁnding solu-

tions, and expressing answers. GIScience education

and geographical education share common goals of

enhancing spatial thinking, using geospatial data, and

preparing students to address real-world challenges.

Efforts and research ﬁndings suggest that GISciences

and geospatial education are more than learning tech-

nical skills but the cognitive ability to solve practi-

cal questions at different stages (Downs et al., 1988;

Verma and Estaville, 2018). However, researchers are

afraid that the use of GISystems as a means will be-

come the end so that the users are not able to think

and solve problems spatially without the help of phys-

ical tools (Downs, 1997). The process is also hard to

scale up considering the relevant cost of money and

time (Johnson and Sullivan, 2010).

On the other hand, the design and usability

of GISystems are improved by making them more

closely resemble the way humans reason about the

world (Goodchild, 2009). User handbook and sup-

porting documentation can be one important facilita-

Characterizing Machine Guidance in Geospatial Analysis

tion. It shows its power and usefulness when users

are new to a certain ﬁeld when timely instructions can

signiﬁcantly improve their analytical process (Ceneda

et al., 2016). The emergence of tutorials and guide-

lines (Perry et al., 2002; Kurland and Gorr, 2007)

offer a means of doing spatial analysis by learning.

Help resources offered by the software vendors (e.g.

ArcGIS Online services) make the GISystems easier

to learn and use (Goodchild, 2000). Implementation

of question-based GISystem (Scheider et al., 2021;

Scheider et al., 2019; Schulze, 2021; Gao and Good-

child, 2013), workﬂow-based GISystem used for an-

alyzing general or domain-speciﬁed problems (Lim

et al., 2005; Yeo and Yee, 2016; Badea and Badea,

2013; Kruiger et al., 2021), and sketch-based GISys-

tem (Curtis, 2012) offered new ways for users in inter-

acting with the GISystems in order to ease analyst’s

difﬁculties when doing geospatial analysis. For in-

stance, users can express spatial relations more efﬁ-

ciently by applying the electronic pen to the map dis-

play (Cohen et al., 1997).

When treating geospatial analysis as an activity,

however, it is an incremental process that involves

a continuous loop of observing and evaluating the

outcomes so that the analytical goal can be reﬁned

(Figure 1). Help is needed from the system’s side

as a mediation tool in reducing the cognitive difﬁ-

culties (Rogers, 2004). Forming the question itself

also involves the efforts in simplifying and divid-

ing the problem into sub-goals and the analytical ap-

proach should actively manage and link a series of

tasks logically (De Smith et al., 2007). Therefore, the

question-based systems and workﬂow-based systems

fail to solve complex spatial analytical problems. Ad-

vancement of the design should go beyond the sim-

ple automation process without the involvement of in-

teractions and feedback on new information. How-

ever, there is still a lack of design in current sys-

tems that support or guide interactive spatial analyt-

ical behaviors. Human ability and skills like commu-

nication and coordination contribute to collaborations

with other people and can be extended to the com-

putational systems in forming human-computer col-

laborations that are the joint efforts of the computer

system(s) and human user(s) towards a shared anal-

ysis goal (Terveen, 1995). One promising research

avenue is to investigate a mixed-initiative approach

that combines system-initiated guidance with user-

initiated guidance to enhance human-machine intel-

ligence (P

erez-Messina et al., 2022).

3 TOWARDS MACHINE

GUIDANCE TO GEOSPATIAL

ANALYSIS

We deﬁne Machine Guidance in geospatial analysis as

a mixed-initiative approach designed to help analysts

navigate complex spatial tasks when they meet difﬁ-

culties. The concept of machine guidance originates

from early automation and industrial control systems,

particularly in manufacturing and robotics (Kendoul,

2012; Li et al., 2009). Over time, ﬁelds such as Visual

Analytics (VA), enabling user-interactive guidance

developed to enhance data visualization and analy-

sis through computer assistance (Ceneda et al., 2016;

Collins et al., 2018). Unlike autonomous agents or

other systems that provide assistance only upon re-

quest, guidance proactively identiﬁes when help is

needed and delivers timely, context-aware support

throughout the analytical workﬂow, enabling analysts

to make well-informed decisions.

Research on conceptualizing the machine guid-

ance design as well as their physical implementa-

tions thrived in VA and data science domains which

have plentiful discussions and years of accumulation

(Ceneda et al., 2016; Ceneda et al., 2018; Ceneda

et al., 2020; P

erez-Messina et al., 2022; Sperrle et al.,

2022). However, the GIS domain has yet to fully for-

malize its integration. There is no formal discussion

on how the guidance approach in guiding geospatial

analysis can be achieved and thus calls for an attribu-

tion. Users of the GIS applications are no longer re-

stricted to the GIS experts but a wider range of users

with expertise in speciﬁc domains (Slocum et al.,

2001). We deﬁne the target users of the machine guid-

ance as those who have occasional needs to conduct

geospatial analytical tasks based on their domain in-

terests and expertise. This group of users are experts

in a particular problem domain but are novices in the

tool domain related to GISystems (Nyerges, 1995).

They are without or with only a little knowledge of

geography or GISceince (Traynor, 1998). More im-

portantly, they have a goal of doing geospatial analy-

sis for critical decision-making.

When supporting geospatial analysis to solve do-

main problems, guidance should assist human ana-

lysts across various analytical stages (P

erez-Messina

et al., 2022). Its essential capability is recognizing

when the user needs help and what kind of help should

be offered at the moment (Ceneda et al., 2016). Fol-

lowed Figure 1, Figure 2 highlighted the moments

when guidance can be introduced. Since geospatial

analysis consists of interconnected sub-tasks and con-

tinuous decision-making, guidance must dynamically

adapt to changes throughout the process. It should

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management

Figure 2: When or where machine guidance can enhance a geospatial analytical workﬂow (outlined in blue).

provide both high-level analytical strategies and com-

putational support, such as selecting appropriate tools

and parameters (Ceneda et al., 2020). Therefore, the

core challenge in designing effective guidance lies in

accurately identifying analytical difﬁculties and de-

livering support with appropriate content.

In the following sections, we ﬁrst identify when

guidance can be needed based on the conceptualized

geospatial analytical process. We then frame and ex-

plain how guidance can be provided in the process. To

demonstrate when and how a human analyst interacts

with and beneﬁts from machine guidance, we present

a hypothetical scenario:

Lucy, a public health promotion analyst, is

working to improve the local food environ-

ment in Wisconsin. Her objective is to iden-

tify counties with limited food access, focus-

ing on those with a high prevalence of food

deserts. Additionally, she aims to prioritize

densely populated counties, as interventions

in these areas would have a greater impact.

With limited state funding, she must strategi-

cally allocate resources to maximize effective-

ness.

Based on this scenario, Table 1 outlines several mo-

ments when guidance can be introduced (MG) along-

side the analyst’s requests and feedback (User).

3.1 When to Provide Guidance

Addressing the question of when guidance is needed,

we propose that timely intervention is essential

whenever analysts face challenges that hinder their

progress. As noted in Figure 2, guidance can be in-

troduced throughout the process from different as-

pects. Particularly, we introduce a conceptual frame-

work that informs the design and development of ma-

chine guidance that operates on three levels:

• Problem-solving Level Guidance helps analysts

spatialize and reﬁne problems. User 1 indi-

cates the analytical goal of understanding the food

shortage conditions so MG 1 provided guidance

by retrieving a dataset representing counties and

food deserts, which offered insights into the cur-

rent food environment in each county. This en-

abled the analyst to develop a foundational un-

derstanding of the phenomenon and prepared the

analyst for subsequent exploratory steps using

the available data. Additionally, analytical goals

may need reﬁnement based on intermediate ﬁnd-

ings. For instance, the initial problem may be ill-

deﬁned or unrealistic. Guidance can assist ana-

lysts in evaluating their goals and making neces-

sary adjustments.

• Operational Level Guidance supports analysts in

developing GIS workﬂows and executing speciﬁc

spatial functions within a GISsystem. When an-

alysts are uncertain about how to tackle a partic-

ular question, guidance can help construct work-

ﬂows by suggesting appropriate methods and op-

erations. Additionally, if analysts are unfamiliar

with speciﬁc tasks, guidance can automate pro-

cesses to abstract operational complexities and

Characterizing Machine Guidance in Geospatial Analysis

Table 1: Interactions between the Lucy (User) and the machine guidance agent (MG) when assisting her in locating targeted

counties.

Dialogue Guidance Response

User 1: I want to ﬁnd counties in

Wisconsin with food shortages.

MG 1: Counties in Wisconsin and the

food deserts in the state are presented.

We found 110 food deserts.

User 2: For each county, I need to

know how many food deserts are there.

MG 2: We counted how many food

deserts by county as a measurement.

User 3: I need to ﬁnd counties with a

large number of food deserts and are

densely populated. How can I do that?

MG 3: We have two strategies. (a)

Manual Selection: You can analyze the

distribution of food desert counts in

relation to population density and

manually select counties that meet

your criteria. While a-1 offers more

ﬂexibility in analyzing each variable

individually, a-2 enhances pattern

recognition by integrating both factors

into a single visual representation,

making it easier to identify areas of

concern at a glance. If the goal is to

detect broad trends and prioritize

interventions efﬁciently, a-2 is the

preferable choice. However, if a more

detailed breakdown of each factor is

required before drawing conclusions,

a-1 may be more useful. (b)

Automated Selection: We have

identiﬁed 6 counties where both the

number of food deserts and the

population density exceed the mean

which may pinpoint areas of interest.

User 4: Strategy (b) results are easier

to follow but I can not deal with all

these counties. How about using more

than 5 food deserts for selection?

MG 4: After adjusting the selection

condition, we found 5 counties.

User 5: I will do more exploration

visually. Thank you.

MG 5: Any other requests? ...

MG 1 MG 2

MG 3 (a)-1 MG 3 (a)-2

MG 3 (b) MG 4

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management

streamline execution. For instance, User 2 explic-

itly expressed the goal of determining the num-

ber of food deserts in each county, prompting the

guidance system to assist by performing the nec-

essary operations, such as a Spatial Join, to di-

rectly achieve this objective (MG 2). This support

alleviated the analyst’s burden of ﬁguring out how

to obtain the counts and which operations to exe-

cute within the GISystem.

• Strategy Level Guidance aids analysts in devis-

ing and selecting problem-solving strategies. In

MG 3, the guidance offers two different strategies

with three different methods. Multiple solutions

may exist for a given problem, requiring analysts

to simulate different processes and explore “what-

if” scenarios (MG 4) to determine the most suit-

able approach. The guidance agent facilitates this

process by assisting analysts in exploring, evaluat-

ing, and comparing different strategies to identify

the optimal solution.

From another perspective, guidance can be initi-

ated either through explicit requests or by inferring

implicit hints. In MG 1, the system proactively as-

sists the analyst by interpreting the phenomenon as a

starting point, fetching data it deems relevant to the

current analytical goal without requiring explicit in-

put. Instead, it infers the speciﬁc data needs based on

the stated objective. In contrast, MG 2 through MG 4

were triggered by the analyst’s direct feedback, where

the guidance responded to explicit requests for assis-

tance regarding their current intention. It ensures that

guidance can take the initiative to help by adapting to

inferred needs as well as meet user-initiated demands,

enhancing human-system collaboration.

3.2 How to Provide Guidance

3.2.1 Guidance in Different Formats

Based on how the guidance is delivered and commu-

nicated to its user, we can categorize the guidance de-

sign based on its output format (P

erez-Messina et al.,

2022). Guidance can be offered with:

• Numeric reports and textual information, like

describing what to do next in words or a list of val-

ues. In our scenario, the guidance provides con-

tinuous textual explanations, enabling the analyst

to clearly understand both the ongoing processes

and the speciﬁc support being offered by the sys-

tem.

• Visual displays, like showing a map, chart, or

diagram on the screen. With the visual clues,

the user does not need to rely on their memo-

ries or imaginations when scoping their questions.

Throughout MG1 to MG4, the guidance system

generates maps as outputs, enabling the analyst

to visualize spatial patterns and facilitating easier

comparisons. This visual evidence improves the

analyst’s ability to track and advance the analyti-

cal process, ensuring it aligns effectively with the

intended goals.

• Physical operations, like direct operation on the

system end to derive the expected output or an-

swer. In MG2, for instance, the Spatial Join oper-

ation was done as part of the guidance behavior.

• Other formats. Other possible formats like

videos, animations, and audio also can be consid-

ered.

The listed formats are not exclusive from each

other and can be combined to provide more details if

needed. For instance, texts and maps are combined to

guide the analyst in our scenario. In MG3, textual ex-

planation offers hints on how to compare and decide

which strategy can be more suitable. Need to men-

tion that incorporating excessive content into a single

guidance—such as using multiple formats, offering

overly detailed information, or combining too many

elements—can lead to cognitive overload. To en-

sure clarity and usability, guidance should avoid over-

whelming analysts with excessive details that may

hinder their ability to make informed judgments. In-

stead, it should prioritize simplicity and focus on de-

livering concise, relevant information tailored to the

user’s immediate needs.

3.2.2 Guidance in Different Degrees

Degree of guidance means how much assistance is

provided (Ceneda et al., 2019; Ceneda et al., 2016).

The guidance that is orienting or directing the user

has a higher degree of freedom and ﬂexibility com-

pared to prescribing guidance that demonstrates a

ﬁxed solution.

• Orienting guidance orients “where to go”. It

only suggests high-level hints and suggestions. It

is aimed at maintaining users’ current thinking

process (“mental map”) and offering the poten-

tial solutions that can be adopted (Ceneda et al.,

2016). At the problem-solving level, it guides

the user to think about and decide on what is the

proper next step.

• Directing guidance directs “what can be cho-

sen”. It offers concrete recommendations like the

possible options and alternatives that can lead to

desired results (Ceneda et al., 2016). It is more

detailed and concrete than the orienting guidance.

In MG3, the guidance provides two strategies in

Characterizing Machine Guidance in Geospatial Analysis

three methods to identify densely populated coun-

ties with a high number of food deserts. By pre-

senting multiple options, it not only offers ﬂexi-

bility but also empowers the analyst to retain con-

trol, allowing them to choose and decide the most

suitable approach for their next steps.

• Prescribing guidance prescribes “what to do”.

It can offer step-by-step directions to solve a

speciﬁc problem. Unlike cases where the ana-

lyst is unsure of the next steps, here the analyst

knows the goal but may not know how to achieve

it. This prescriptive process can largely be au-

tomated (Ceneda et al., 2016). For instance, the

process of calculating counts (MG2), and the task

of applying a different selection condition (MG4),

are both automated. This approach conceals the

operational complexities and presents only the ﬁ-

nal outputs.

3.3 Design Challenges and

Opportunities

Designing a computational system capable of provid-

ing effective machine guidance for spatial problem-

solving presents several challenges. One of the pri-

mary difﬁculties lies in recognizing when and how to

intervene without disrupting the analyst’s workﬂow.

The assistance should only be provided when it is

genuinely needed and when the analyst is ready to re-

ceive it (Maes, 1995). Otherwise, unnecessary inter-

ruptions could confuse the analyst and interfere with

the analytical process (Ceneda et al., 2020). Detect-

ing the need for assistance can be achieved through

explicit user requests or implicit behavioral moni-

toring, such as tracking user interactions, detecting

recurring difﬁculties, or even analyzing physiolog-

ical indicators like stress-related facial expressions

(Ceneda et al., 2016; Ceneda et al., 2021). A well-

designed system must balance responsiveness with

non-intrusiveness to ensure a seamless analytical ex-

perience (Maes, 1995).

Another signiﬁcant computational challenge in-

volves equipping the system with reasoning and plan-

ning capabilities. The system must be able to track

the progress of an analysis, recognize what has been

completed, identify gaps in the current strategy, and

suggest appropriate next steps. This requires integrat-

ing structural knowledge, which enables the system

to monitor and manage analytical workﬂows (Arm-

strong et al., 1990). Furthermore, a control process

is needed to oversee and guide execution, ensuring

that if an approach fails, alternative solutions can be

explored (Hayes-Roth and Hayes-Roth, 1979). The

system must also infer the analyst’s current stage and

anticipate their information needs, adjusting guidance

accordingly. Recent advancements in Artiﬁcial In-

telligence (AI), particularly Large Language Mod-

els (LLMs) and GeoAI, offer promising solutions for

enhancing the design of machine guidance agents.

These technologies can improve the agent’s ability

to interpret user queries, infer analytical goals, and

adaptively learn from user interactions to reﬁne fu-

ture assistance (Li and Ning, 2023). A learning mod-

ule could be integrated to analyze user behavior over

time, helping the system provide more personalized

and context-aware recommendations (Smith et al.,

1987). Leveraging AI-driven techniques can signif-

icantly enhance the effectiveness of computational

guidance systems by making them more intuitive,

adaptive, and capable of supporting complex spatial

analyses.

How to store and represent the required expertise

is another essential challenge. The mentioned ex-

pertise should encompass various types of informa-

tion, including declarative knowledge (basic informa-

tion about spatial and non-spatial objects), procedu-

ral knowledge (strategies for applying spatial opera-

tions), and control knowledge (heuristics for evaluat-

ing the validity of actions and solutions) (Hofer et al.,

2017). The system must be able to intelligently select

and compare spatial operations based on their pur-

pose and suitability for different analytical goals. For

instance, when selecting datasets for Wisconsin, the

representation could vary based on the scale and pur-

pose of analysis, whether as a point on a small-scale

map or as a polygon in a larger-scale map. Knowl-

edge of metadata such as geometry type, spatial ex-

tent, and scale is crucial for ensuring that the correct

datasets are retrieved and applied computationally.

4 DISCUSSION

To ensure the successful implementation of the ma-

chine guidance approach, several critical research

questions must be addressed. First, the conceptual

framework requires further evaluation and reﬁnement.

While our characterizations (Sections 3.1, 3.2) pro-

vide an initial understanding of the machine guidance

approach in geospatial analysis, it remains uncertain

whether they accurately reﬂect real-world analytical

workﬂows, the speciﬁc challenges analysts face, and

the most effective formats for delivering guidance.

One way to bridge this gap is through observational

studies to examine how analysts approach geospatial

problem-solving in real-world scenarios. Addition-

ally, empirical research is needed to determine the

optimal timing and presentation of guidance to maxi-

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management

mize its effectiveness. A user-centered approach will

offer valuable insights for designing more intuitive

and effective machine-guided systems.

Secondly, it is computationally challenging to de-

termine when to provide the guidance. Although we

have outlined potential strategies in Section 3.3, their

implementation is still limited, particularly in the con-

text of geospatial analysis. Future research should

explore methods to intelligently recognize the need

for guidance, ensuring that it is provided in a timely

and non-disruptive manner. This could involve devel-

oping adaptive systems that learn from user behavior

and context to offer guidance only when it is most

beneﬁcial.

Third, ensuring the scalability of the expertise

embedded within the guidance while maintaining its

speciﬁcity across diverse domains is critical. This

raises important questions about what information or

knowledge should be captured, how it should be rep-

resented, and the level of detail required to balance

generality and domain-speciﬁc relevance. Addressing

these questions is essential for creating guidance sys-

tems that are both adaptable to a wide range of appli-

cations and sufﬁciently detailed to provide meaning-

ful support in specialized contexts. Future research

should explore methods for capturing and structuring

domain knowledge in a way that ensures scalability

without sacriﬁcing precision and ultimately ensures

machine guidance applications are effective across

various analytical domains.

To sum up, our study is to establish proof-

of-concept design principles for machine guidance.

While we have identiﬁed various design opportu-

nities, our work remains at the conceptual stage

and lacks computational implementation and prac-

tical evaluation. Future efforts are needed to de-

velop concrete computational frameworks for ma-

chine guidance and to conduct empirical evaluations.

Implementing these systems in real-world applica-

tions will provide valuable insights through user feed-

back, which can help reﬁne and validate the proposed

design features. Addressing the mentioned challenges

will not only advance the ﬁeld of machine-guided

geoanalytical systems but also pave the way for

broader applications in other domains where human-

machine collaboration is critical.

5 CONCLUSIONS

We propose using machine guidance as a new ap-

proach to support analysts’ geospatial analytical pro-

cess. As a complex and mentally demanding en-

deavor, geospatial analysis poses a signiﬁcant chal-

lenge for individuals with limited GIS knowledge and

experience. This challenge highlights the importance

of machine guidance, which encapsulates reasoning

and processing steps within the system itself, allow-

ing analysts to focus on interpreting analytical outputs

rather than navigating complex workﬂows.

As a starting point, we characterized a conceptual

framework focusing on (1) when guidance should be

introduced and (2) how guidance can be delivered to

help the analyst. Speciﬁcally, our approach suggests

that machine guidance should be placed when the user

meets difﬁculties when doing geospatial analysis. At

these moments, users may struggle with decision-

making due to the need for extensive mental effort, of-

ten lacking sufﬁcient knowledge or expertise to solve

the issue at hand. Using the hypothetical scenario, we

presented various machine guidance that are in differ-

ent degrees and formats.

To the best of our knowledge, this research marks

the ﬁrst attempt to integrate machine guidance into

GISystems, with the potential to drive advancements

in the broader ﬁeld of GIScience. As an initial ef-

fort, this study introduces the concept of the machine

guidance approach in supporting geospatial analysis

as well as its design opportunities and potential. Our

ongoing efforts are focused on reﬁning these guid-

ance design strategies and evaluating their feasibil-

ity in real-world scenarios. We seek to foster further

academic discussion and social engagement around

the role of machine guidance in facilitating geospa-

tial analysis. Ultimately, we hope this work sparks

further investigation and encourages the development

of innovative approaches along the way.

REFERENCES

Ahasan, R. and Hossain, M. M. (2021). Leveraging gis

and spatial analysis for informed decision-making in

covid-19 pandemic. Health policy and technology,

10(1):7.

Armstrong, M., De, S., Densham, P., Lolonis, P., Rushton,

G., and Tewari, V. K. (1990). A knowledge-based

approach for supporting locational decisionmaking.

Environment and Planning B: Planning and Design,

17(3):341–364.

Baars, B. J. (1997). Some essential differences between

consciousness and attention, perception, and working

memory. Consciousness and cognition, 6(2-3):363–

371.

Baddeley, A. (1992). Working memory: The interface be-

tween memory and cognition. Journal of cognitive

neuroscience, 4(3):281–288.

Badea, A.-C. and Badea, G. (2013). The advantages of

creating compound gis functions for automated work-

Characterizing Machine Guidance in Geospatial Analysis

ﬂow. International Multidisciplinary Scientiﬁc Geo-

Conference: SGEM, 1:943.

Bailey, T. C., Gatrell, A. C., et al. (1995). Interactive spa-

tial data analysis, volume 413. Longman Scientiﬁc &

Technical Essex.

Bednarz, S., Heffron, S., and Huynh, N. (2013). A road map

for 21st century geography education. Washington,

DC: Association of American Geographers.

Bednarz, S. W. (2004). Geographic information systems: A

tool to support geography and environmental educa-

tion? GeoJournal, 60:191–199.

Ceneda, D., Andrienko, N., Andrienko, G., Gschwandtner,

T., Miksch, S., Piccolotto, N., Schreck, T., Streit, M.,

Suschnigg, J., and Tominski, C. (2020). Guide me

in analysis: A framework for guidance designers. In

Computer Graphics Forum, volume 39, pages 269–

288. Wiley Online Library.

Ceneda, D., Arleo, A., Gschwandtner, T., and Miksch, S.

(2021). Show me your face: towards an automated

method to provide timely guidance in visual analyt-

ics. IEEE Transactions on Visualization and Com-

puter Graphics, 28(12):4570–4581.

Ceneda, D., Gschwandtner, T., May, T., Miksch, S., Schulz,

H.-J., Streit, M., and Tominski, C. (2016). Character-

izing guidance in visual analytics. IEEE transactions

on visualization and computer graphics, 23(1):111–

120.

Ceneda, D., Gschwandtner, T., May, T., Miksch, S., Streit,

M., and Tominski, C. (2018). Guidance or no guid-

ance? a decision tree can help. In EuroVA@ EuroVis,

pages 19–23.

Ceneda, D., Gschwandtner, T., and Miksch, S. (2019). A

review of guidance approaches in visual data analy-

sis: A multifocal perspective. In Computer Graphics

Forum, volume 38, pages 861–879. Wiley Online Li-

brary.

Cohen, P. R., Johnston, M., McGee, D., Oviatt, S., Pittman,

J., Smith, I., Chen, L., and Clow, J. (1997). Quickset:

Multimodal interaction for distributed applications. In

Proceedings of the ﬁfth ACM international conference

on Multimedia, pages 31–40.

Collins, C., Andrienko, N., Schreck, T., Yang, J., Choo, J.,

Engelke, U., Jena, A., and Dwyer, T. (2018). Guid-

ance in the human–machine analytics process. Visual

Informatics, 2(3):166–180.

Couclelis, H. (2003). The certainty of uncertainty: Gis and

the limits of geographic knowledge. Transactions in

GIS, 7(2):165–175.

Council, N. R. et al. (2005). Learning to think spatially.

Creem-Regehr, S. H., Willemsen, P., Gooch, A. A., and

Thompson, W. B. (2005). The inﬂuence of restricted

viewing conditions on egocentric distance perception:

Implications for real and virtual indoor environments.

Perception, 34(2):191–204.

Curtis, J. W. (2012). Integrating sketch maps with gis to ex-

plore fear of crime in the urban environment: A review

of the past and prospects for the future. Cartography

and Geographic Information Science, 39(4):175–186.

De Smith, M. J., Goodchild, M. F., and Longley, P. (2007).

Geospatial analysis: a comprehensive guide to prin-

ciples, techniques and software tools. Troubador pub-

lishing ltd.

orner, D. and Kreuzig, H. W. (1983). Prob-

leml

osef

ahigkeit und intelligenz. Psychologische

Rundschau.

Downs, R. M. (1997). The geographic eye: Seeing through

gis? 1. Transactions in GIS, 2(2):111–121.

Downs, R. M., Liben, L. S., and Daggs, D. G. (1988).

On education and geographers: The role of cogni-

tive developmental theory in geographic education.

Annals of the Association of American Geographers,

78(4):680–700.

Dramowicz, K., Wightman, J. F., and Crant, J. S. (1993).

Addressing gis personnel requirements: A model for

education and training. Computers, Environment and

Urban Systems, 17(1):49–59.

Dungan, J. L., Perry, J., Dale, M., Legendre, P., Citron-

Pousty, S., Fortin, M.-J., Jakomulska, A., Miriti, M.,

and Rosenberg, M. (2002). A balanced view of scale

in spatial statistical analysis. Ecography, 25(5):626–

640.

Fischer, A., Greiff, S., and Funke, J. (2011). The process of

solving complex problems. Journal of Problem Solv-

ing, 4(1):19–42.

Freksa, C., Barkowsky, T., Dylla, F., Falomir, Z., Oltet¸eanu,

A.-M., and van de Ven, J. (2017). Spatial problem

solving and cognition. In Representations in mind and

world, pages 156–183. Routledge.

Gao, S. and Goodchild, M. F. (2013). Asking spatial ques-

tions to identify gis functionality. In 2013 Fourth In-

ternational Conference on Computing for Geospatial

Research and Application, pages 106–110. IEEE.

Goodchild, M. and Longley, P. (1999). The future of gis and

spatial analysis. Geographical information systems,

1:567–580.

Goodchild, M. F. (1992). Geographical information sci-

ence. International journal of geographical informa-

tion systems, 6(1):31–45.

Goodchild, M. F. (2000). Part 1 spatial analysts and gis

practitioners: The current status of gis and spatial

analysis. Journal of Geographical Systems, 2:5–10.

Goodchild, M. F. (2006). The fourth r? rethinking gis edu-

cation. ESRI ArcNews, 28(3):1.

Goodchild, M. F. (2009). Geographic information systems

and science: Today and tomorrow. Annals of GIS,

15(1):3–9.

Goodchild, M. F., Anselin, L., Appelbaum, R. P., and

Harthorn, B. H. (2000). Toward spatially integrated

social science. International Regional Science Re-

view, 23(2):139–159.

Goodchild, M. F. and Janelle, D. G. (2010). Toward critical

spatial thinking in the social sciences and humanities.

GeoJournal, 75:3–13.

Grekousis, G. (2020). Spatial analysis methods and prac-

tice: describe–explore–explain through GIS. Cam-

bridge University Press.

Haining, R. (1994). Designing spatial data analysis modules

for geographical information systems. Spatial analy-

sis and GIS, pages 45–64.

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management

Hayes-Roth, B. and Hayes-Roth, F. (1979). A cognitive

model of planning. Cognitive science, 3(4):275–310.

Hofer, B., M

as, S., Brauner, J., and Bernard, L. (2017).

Towards a knowledge base to support geoprocessing

workﬂow development. International Journal of Ge-

ographical Information Science, 31(4):694–716.

Huisman, O., de By, R. A., et al. (2009). Principles of ge-

ographic information systems. ITC Educational Text-

book Series, 1:17.

Johnson, A. B. and Sullivan, D. (2010). Geospatial edu-

cation at us community colleges: Background, chal-

lenges, and opportunities. Journal of the Urban &

Regional Information Systems Association, 22(2).

Kang, J.-Y., Michels, A., Lyu, F., Wang, S., Agbodo, N.,

Freeman, V. L., and Wang, S. (2020). Rapidly mea-

suring spatial accessibility of covid-19 healthcare re-

sources: a case study of illinois, usa. International

journal of health geographics, 19:1–17.

Kendoul, F. (2012). Survey of advances in guidance, nav-

igation, and control of unmanned rotorcraft systems.

Journal of Field Robotics, 29(2):315–378.

Kruiger, J. F., Kasalica, V., Meerlo, R., Lamprecht, A.-L.,

Nyamsuren, E., and Scheider, S. (2021). Loose pro-

gramming of gis workﬂows with geo-analytical con-

cepts. Transactions in GIS, 25(1):424–449.

Kurland, K. S. and Gorr, W. L. (2007). GIS tutorial for

health. ESRI, Inc.

Li, M., Imou, K., Wakabayashi, K., and Yokoyama, S.

(2009). Review of research on agricultural vehicle au-

tonomous guidance. International Journal of Agricul-

tural and biological engineering, 2(3):1–16.

Li, Z. and Ning, H. (2023). Autonomous gis: the next-

generation ai-powered gis. International Journal of

Digital Earth, 16(2):4668–4686.

Lieberman, H., Fry, C., and Rosenzweig, E. (2015). The

New Era of High-Functionality Interfaces. In Duval,

B., Van Den Herik, J., Loiseau, S., and Filipe, J., edi-

tors, Agents and Artiﬁcial Intelligence, volume LNAI

8946, pages 3–10. Springer International Publishing,

Cham.

Lim, K. J., Engel, B. A., Tang, Z., Choi, J., Kim, K.-S.,

Muthukrishnan, S., and Tripathy, D. (2005). Auto-

mated web gis based hydrograph analysis tool, what

1. JAWRA Journal of the American Water Resources

Association, 41(6):1407–1416.

Liu, S., Andrienko, G., Wu, Y., Cao, N., Jiang, L., Shi, C.,

Wang, Y.-S., and Hong, S. (2018). Steering data qual-

ity with visual analytics: The complexity challenge.

Visual Informatics, 2(4):191–197.

Maes, P. (1995). Agents that reduce work and information

overload. In Readings in human–computer interac-

tion, pages 811–821. Elsevier.

Miller, H. J. and Wentz, E. A. (2003). Representation and

spatial analysis in geographic information systems.

Annals of the Association of American Geographers,

93(3):574–594.

Mollalo, A., Mohammadi, A., Mavaddati, S., and Kiani, B.

(2021). Spatial analysis of covid-19 vaccination: a

scoping review. International journal of environmen-

tal research and public health, 18(22):12024.

Nyerges, T. L. (1995). Cognitive issues in the evolution of

gis user knowledge. In Cognitive aspects of human-

computer interaction for geographic information sys-

tems, pages 61–74. Springer.

erez-Messina, I., Ceneda, D., El-Assady, M., Miksch, S.,

and Sperrle, F. (2022). A typology of guidance tasks

in mixed-initiative visual analytics environments. In

Computer Graphics Forum, volume 41, pages 465–

476. Wiley Online Library.

Perry, J., Liebhold, A., Rosenberg, M., Dungan, J., Mir-

iti, M., Jakomulska, A., and Citron-Pousty, S. (2002).

Illustrations and guidelines for selecting statistical

methods for quantifying spatial pattern in ecological

data. Ecography, 25(5):578–600.

Pretz, J. E., Naples, A. J., and Sternberg, R. J. (2003). Rec-

ognizing, deﬁning, and representing problems. The

psychology of problem solving, 30(3):3–30.

Ram

ırez, I. J. and Lee, J. (2020). Covid-19 emergence and

social and health determinants in colorado: a rapid

spatial analysis. International journal of environmen-

tal research and public health, 17(11):3856.

Roche, S. (2014). Geographic information science i:

Why does a smart city need to be spatially enabled?

Progress in Human Geography, 38(5):703–711.

Rogers, Y. (2004). New theoretical approaches for hci. An-

nual review of information science and technology,

38(1):87–143.

Scheider, S., Ballatore, A., and Lemmens, R. (2019). Find-

ing and sharing gis methods based on the questions

they answer. International journal of digital earth,

12(5):594–613.

Scheider, S., Nyamsuren, E., Kruiger, H., and Xu, H.

(2021). Geo-analytical question-answering with gis.

International Journal of Digital Earth, 14(1):1–14.

Schulze, U. (2021). “gis works!”—but why, how, and for

whom? ﬁndings from a systematic review. Transac-

tions in GIS, 25(2):768–804.

Shin, E. E., Milson, A. J., and Smith, T. J. (2016). Future

teachers’ spatial thinking skills and attitudes. Journal

of Geography, 115(4):139–146.

Slocum, T. A., Blok, C., Jiang, B., Koussoulakou, A.,

Montello, D. R., Fuhrmann, S., and Hedley, N. R.

(2001). Cognitive and usability issues in geovisual-

ization. Cartography and geographic information sci-

ence, 28(1):61–75.

Smith, T., Peuquet, D., Menon, S., and Agarwal, P. (1987).

Kbgis-ii a knowledge-based geographical information

system. International Journal of Geographical Infor-

mation System, 1(2):149–172.

Sperrle, F., Ceneda, D., and El-Assady, M. (2022). Lotse:

A practical framework for guidance in visual analyt-

ics. IEEE Transactions on Visualization and Com-

puter Graphics, 29(1):1124–1134.

Sun, F., Matthews, S. A., Yang, T.-C., and Hu, M.-H.

(2020). A spatial analysis of the covid-19 period

prevalence in us counties through june 28, 2020:

where geography matters? Annals of epidemiology,

52:54–59.

Terveen, L. G. (1995). Overview of human-computer col-

laboration. Knowledge-Based Systems, 8(2-3):67–81.

Characterizing Machine Guidance in Geospatial Analysis

Thomas, J. J. (2005). Illuminating the path:the research

and development agenda for visual analytics. IEEE

Computer Society.

Traynor, C. (1998). Putting power in the hands of end users:

a study of programming by demonstration, with an ap-

plication to geographical information systems. In CHI

98 conference summary on Human factors in comput-

ing systems, pages 68–69.

Verma, K. and Estaville, L. (2018). Role of geography

courses in improving geospatial thinking of under-

graduates in the united states. International Journal

of Geospatial and Environmental Research, 5(3):2.

Wang, Q., Chen, Z., Wang, Y., and Qu, H. (2021). A sur-

vey on ml4vis: Applying machinelearning advances

to data visualization. IEEE Transactions on Visual-

ization and Computer Graphics.

Ward, T. B., Smith, S. M., and Finke, R. A. (1999). Creative

cognition. Handbook of creativity, 189:212.

Yeo, I.-A. and Yee, J.-J. (2016). Development of an auto-

mated modeler of environment and energy geographic

information (e-gis) for ecofriendly city planning. Au-

tomation in Construction, 71:398–413.

GISTAM 2025 - 11th International Conference on Geographical Information Systems Theory, Applications and Management