ABOUT

The Idea

Have you ever wondered how data about our world is generated? How can we know what people think about different environments and how people behave in certain places? One way of finding out is by directly asking people in the field. However, to find out common trends, a lot of data is needed and directly asking enough people becomes impossible. Another way is to use online tools such as questionnaires and surveys to ask many people at once. However, let’s be honest, these are rarely any fun and most people aren’t eagerly waiting for a new online survey to come out. So what can we offer people so that they will want to contribute scientific data and have fun doing so…?

You have probably heard of the worldwide phenomenon Pokemon GO by now, maybe you have also heard of similar types of entertainment such as INGRESS or GeoCaching? These location-based games are played by millions of people and generate enormous amounts of data about where people like to go, what people like to do and the things that are important to people.

Arcane Shift is inspired by these projects and wants to offer you a chance to contribute scientific data that will actually be used in science, all whilst having fun playing a location-based game. 

But what is Arcane Shift all about? What kind of game am I making? The idea of Arcane Shift is best described as a combination of Pokemon GO, Settlers of Catan and Risk with elements of popular fantasy games such as World of Warcraft and Rust. Since I am developing the game mostly by myself (and game development takes its time) I will be developing the game in three main chapters. If you have any great ideas or if you want to get involved and contribute to the project, feel free to drop a line in the community forum. 

What is Arcane Shift?

Arcane Shift is a location-based game that will be developed as part of an ongoing PostDoc research project in gamification, active crowdsourcing and landscape perception research. The game aims to unite entertainment and high quality data generation.

For effective landscape research, high quality underlying data from a diverse crowd is crucial. However, both traditional expert-based methods as well as modern participatory approaches face several limitations. Traditional methods of generating landscape-relevant information are confined by financial and time limitations, whereas modern crowdsourcing and citizen science approaches often struggle with user engagement, retention, and bias. Gamification – adding entertaining elements to a process – has been found to encourage a diverse audience, increase user motivation and retention, produce high quality data for academic research and has successfully been incorporated into many crowdsourcing efforts.

This interdisciplinary research project will transcend contemporary crowdsourcing efforts by exploiting the unique motivational and spatial characteristics of a novel location-based game to generate rich landscape perception and preference data. The generated data will be analysed using methods from geographic information science, computational linguistics, and machine learning to extract and explore emergent landscape dimensions. 

Detailed Description

Summary

Landscapes actualise our existence: they are the perceivable areas of the earth in which our lives unfold, providing tangible (e.g. food, water, building materials) and intangible (e.g. recreational potential, aesthetic value, inspiration) dimensions to our lives. They influence our social, physical and mental well-being (Abraham et al., 2010) and their importance is reflected in emergent frameworks and international conventions. Our perceptions of and interactions with our surroundings has been the topic of interest in landscape perception and preference research for which high quality data sets capturing how people perceive their surroundings are crucial. This is of particular importance seeing the detailed insights into individual and shared values of landscapes such data can provide to support fair decision and inclusive policy making (Scott, 2003). Traditional methods of generating landscape-relevant information are confined by costs as well as available time and modern crowdsourcing and citizen science approaches often struggle with user engagement, retention and bias. Gamification – adding entertaining elements to a process – has been found to increase user motivation and retention, produce heterogeneous data for academic research and has successfully been incorporated into many crowdsourcing efforts (Morschheuser et al., 2017). However, even though complementing landscape perception data generation with gamified elements has explicitly been called for (Bubalo et al., 2019), projects remain scarce. My interdisciplinary research will transcend contemporary crowdsourcing efforts by exploiting the unique motivational and spatial characteristics of a novel location based game to generate rich landscape perception and preference data. I will develop, implement, test and run a globally accessible location based game built using real-world data, to motivate users to contribute landscape relevant information. The generated data will be analysed using methods from geographic information science as well as computational linguistics and machine learning to extract and explore relevant landscape dimensions. The results are expected to showcase the potential of gamification in academic data generation as well as allow for the quantification and qualitative characterisation of landscape perceptions within and between local and visiting individuals as well as socio-demographic and cultural groups. This has important implications for both participatory data generation in science as well as landscape policy and decision making. By evaluating the developed application I will test the hypothesis that location based games offer a powerful tool for environmental, experiential and perceptual data generation. Findings will be published as a set of recommendations based on descriptive statistics derived from the implemented application. Further, extracting salient landscape information from the generated data will allow the compilation of important perceived interactions and dimensions, which will be published to complement future inquiries into landscapes. In addition, the assumption of socio-demographic and cultural differences in landscape perceptions as well as local and visiting individuals’ values will be investigated and significant similarities and differences will be reported, contributing to debates of inclusion and sustainability in regards to landscapes. Seeing the participatory nature of the research, the results will be communicated through scientific articles as well as public events to further the academic discourse as well as educate and empower participating individuals. Ultimately, this research will support important societal questions of how to reconcile conserving valuable landscape dimensions with sustaining economic operations (e.g. forestry, recreation, hydroelectricity, tourism) in New Zealand and transferable to Switzerland or any region of interest where the proposed location based game is enjoyed.

Current state of research in the field

Landscapes have inspired laypeople, artists and scientists for centuries, from beautiful landscape paintings, over vivid descriptions in books and poems, to explorations of landscape perception and preference (Antrop, 2013). Contemporary research is predominantly interested in how individuals or specific groups of individuals perceive landscapes (Scott, 2003; Zube et al., 1982), how landscapes influence mental, physical and social well-being (Abraham et al., 2010; Thompson Coon et al., 2011), what dimensions of landscapes are particularly salient (van Putten et al., 2020; Zube et al., 1982) and how landscapes influence our behaviour (Heft, 2010). These questions are gaining prominence in an interconnected and increasingly urbanised world to inform inclusive future policies and decisions about our environment.

Traditionally, research on landscapes has primarily been based on expert opinions and the dichotomy between laypersons and experts was neglected (Lowenthal & Prince, 1965; Swanwick, 2009). However, the introduction of the European Landscape Convention (ELC) and their widely adopted definition of landscape as “an area, as perceived by people, whose character is the result of the action and interaction of natural and/or human factors” (European Landscape Convention, 2000, p. 2) called for a more participatory approach. This gradually led to a paradigm shift from top-down expert based research to a more bottom-up participatory approach of acknowledging and including lay people’s opinions and local knowledge. Especially prominent and widely adopted are the two approaches Ecosystem Services (ES) (Millennium Ecosystem Assessment (MEA), 2005) and Landscape Character Assessment (LCA) (Tudor, 2014). These differ in their focus with the LCA shedding light on the experiential and perceptual dimensions of landscapes whereas the ES approach focuses more on the biophysical dimensions and the values of these, usually in monetary terms. A further important albeit less common approach is exploring landscapes through their affordances (possibilities of action) (Heft, 2010), which have been argued to be particularly suited to investigate temporal aspects of how landscapes and places are perceived (Raymond et al., 2017). All three mentioned approaches have complemented landscape perception research and have led to a deeper understanding of the importance of landscapes and how we interact with these.

To delve into questions of landscape perception, underlying datasets are essential, commonly consisting of preference ratings, elicited points or areas of interest and various corpora of natural language. Ratings of landscape scenes and images allow for inquiries into landscape appreciation and scenicness (Byoung-Eyang & Kaplan, 1990; Kaplan & Herbert, 1987; Petrova et al., 2015; Seresinhe et al., 2018). Points or areas of interest, frequently collected through participatory approaches such as public participation geographic information systems (PPGIS) (Brown et al., 2020), provide insights into salient locations as well as landscape elements and how these are perceived and valued (Alessa et al., 2008; Plieninger et al., 2013; Solecka et al., 2022). Natural language corpora collected through surveys and questionnaires (Fagerholm et al., 2020; Hedblom et al., 2020), free-listing experiments (Wartmann et al., 2018), image descriptions (Wartmann, Koblet, et al., 2021) and short stories (Bieling, 2014) have been found to capture various dimensions of a landscape’s character and have been used to explore intangible dimensions of a given landscape. However, mentioned datasets are accompanied by various limitations, most prominent of which are the confounding factors of limited time and finances to conduct large-scale moderated participant sessions. Recently, inquiries into how we interact with and perceive our surroundings have been complemented with social media or otherwise crowdsourced data (Chen et al., 2018; Huang et al., 2013; Wartmann, Baer, et al., 2021), offering a cost effective approach to generating a large amount of data (See et al., 2016). However, user generated content is commonly noisy, requiring complex methods of identifying relevant documents as well as being biassed towards certain socio-demographic groups (Ghermandi & Sinclair, 2019)

Including an often neglected audience of younger participants through a more playful approach has been called for, specifically in landscape perception and preference research (Bubalo et al., 2019). Gamification refers to “hedonic or entertainment-oriented technologies being re-appropriated for productive use” (Koivisto & Hamari, 2019, p. 191) or in other words complementing an existing process with entertaining or playful elements (Hamari et al., 2014). The literature shows gamification successfully increases user motivation as well as retention and has primarily been incorporated in educational and crowdsourcing efforts (Hamari et al., 2014; Morschheuser et al., 2017, 2019). Spatial gamified applications are plentiful in academia as well as the private sector and found on a gradient between more serious and goal oriented applications such as crowdsourcing hydrology data as in CrowdWater (Strobl et al., 2019), to more ludic and entertaining applications such as Ingress and Pokemon GO collecting points of interest (Laato et al., 2019). Further examples related to geographic information science include, but are not limited to, Feeding Yoshi, Urbanopoly, Cropland Capture, FotoQuest Austria and Sea Hero Quest. Feeding Yoshi, an early example of a location based game (LBG), which successfully collected information on available WiFi networks (Bell et al., 2006) and highlighted the potential of LBGs to motivate participation in scientific studies. Urbanopoly was able to incorporate and generate open and linked geo-spatial data in urban environments through location based gameplay (Celino et al., 2012), suggesting that real-world data can not only be used to build a virtual game world, but can also be curated through a playful application. Cropland Capture and FotoQuest Austria revolve around land cover data generation and validation with Cropland Capture motivating participants to judge the presence or absence of cropland in defined areas (Salk et al., 2016) and FotoQuest Austria being a LBG to generate detailed in-situ land cover reports in predefined locations (Bayas et al., 2016). These applications go to show the potential of using gamification and LBGs to generate information about how people perceive their environment, albeit through predefined land cover categories. Lastly, Sea Hero Quest is an example of a highly gamified application in which players were tasked with navigating a virtual game world. This academic game successfully motivated millions of participants leading to unprecedented explorations of spatial cognition and navigational skills as well as uncovering correlations with dementia (Coutrot et al., 2018). The sheer number of participants and the quality of the generated data and results proves the power of well-designed gamified applications in generating unmatched quantities of data, leading to new avenues of research. Of these, location based games provide a particularly promising platform to explore how users experience and perceive their surroundings, especially since moving around in the real world is integrated as part of the gameplay (Neustaedter et al., 2013). This allows for in-situ volunteered geographic information generation and exploiting individuals as sensors capable of providing rich experiential and perceptual data (Goodchild, 2007). In addition, location based games are highly adaptable and allow for the collection of various forms of data, from images and texts such as in Ingress, over location based information such as in Feeding Yoshi, to directional images and descriptions of land cover as in FotoQuest Austria. Future data needs can easily be incorporated into the gameplay resulting in a sustainable and modular landscape relevant data generation platform. 

Despite the mentioned advantages of gamifying data generation approaches, gamification has rarely been incorporated into landscape perception data generation beyond land cover and land use judgements. Further, active crowdsourcing platforms on the more playful end of the gamification gradient (e.g. location based games) are nigh inexistant. This baffling discrepancy between the literature agreeing that gamification is a powerful tool of increasing user motivation, user retention as well as data quality and the lack of research on gamifying landscape perception data generation outlines major research gaps which I aim to address. I have formulated the following overarching research questions to be addressed in this research fellowship:

  • ORQ1: How can a location based game be developed and implemented to actively crowdsource in-situ landscape perception and preference information?
  • ORQ2: How does landscape perception and preference data crowdsourced through a gamified application compare to traditional approaches of data generation, particularly authoritative datasets, in terms of accuracy, credibility, and overall quality?
  • ORQ3: What are similarities and differences in how landscapes are perceived within and between socio-demographic and cultural groups as well as local and visiting individuals and what are the implications for landscape characterisations as well as policy and decision makers?

Current state of my own research

To explore the potential of gamification in landscape perception and preference research, I developed and implemented a LBG called StarBorn to crowdsource in-situ land cover information (Baer et al., 2019). The implemented LBG successfully generated over 13’000 contributions in a period of around three months, underlining the potential of LBGs to generate large amounts of data. The results show interesting discrepancies between an authoritative land cover dataset and user contributions, pointing towards differences in how specific areas are perceived (e.g. confusions between urban green area and forest). In addition, a negative correlation was found between the number of contributions per user and respective contribution quality, calling for particular consideration of participants maximising in-game performance over prioritising data quality. In a further study we investigated the potential of using passively crowdsourced social media data to estimate the recreational potential of areas in Switzerland (Wartmann, Baer, et al., 2021). The results again show differences between authoritative datasets and user generated content. In particular, areas of low recreational potential according to the authoritative dataset, but showing high recreational potential according to the social media dataset were identified. 

When starting my PhD, we envisioned a project revolving around a location based game to collect experiential and perceptual data on how individuals interact with their surroundings. However, due to the global covid-19 pandemic and the entailed restrictions on social gatherings as well as the reduced mobility through various forms of lockdown, this initial idea was abandoned in favour of a safer, albeit less gamified approach. However, we did explore how gamification crystalised around emergent borders and compiled a list of examples (Thibault & Baer, 2021). In parallel, I developed a web-application from scratch with the aim of crowdsourcing natural language descriptions of people’s everyday lived landscapes, whilst allowing participation safely from home (Thibault & Baer, 2021). The implemented platform, adequately named Window Expeditions, sought to capitalise on people’s reduced mobility whilst offering an opportunity to explore the world through written descriptions. Window Expeditions was able to generate around 650 natural language descriptions rich in landscape relevant information in three languages. The results show that the contributed descriptions indeed capture salient landscape dimensions of everyday lived landscapes of the participants and that natural language allows for detailed explorations of underlying semantics. Window Expeditions successfully captured various biophysical elements as well as sensory experiences and cultural ecosystem services of the immediate surroundings of participating users. However, the generated dataset is rather small and thus questions of user motivation and retention remain important.

My academic activities, and particularly my unique experiences in implementing various (gamified) platforms and analysing the contributed data, leave me in a pole position to continue establishing gamification in landscape perception and preference research. My passion for my initially envisioned PhD project of creating and analysing a location based game for landscape perception and preference research remains unwavered and a new post-pandemic normal slowly breathing life into the world presents a unique opportunity of starting this research.

Detailed Research Plan

Using scientific frameworks and guided by findings in the literature, lessons learnt from previous implementations and the expertise of the host supervisor and institute, I will develop and evaluate a software prototype of a location based game geared towards landscape relevant data generation. The application will be opened to the public and promoted through various channels to ensure a large and diverse audience. I will first explore descriptive statistics derived from the implemented application to guide future participatory data generation efforts. Further, salient landscape information will be extracted and collated to a collection of experiential and perceptual landscape dimensions to support policy and decision making processes. Finally, socio-demographic and cultural differences in landscape perceptions as well as differences between local and visiting individuals’ perceptions will be investigated and significant similarities and differences will be elicited, contributing to important societal debates in regards to landscapes and how we interact with these.

Developing a novel LBG for landscape perception and preference research

Using a scientifically based requirements elicitation framework (Carrizo et al., 2014) in combination with common gamification frameworks (Hunicke et al., 2004; Martella et al., 2015) and the technical as well as infrastructural support of the Centre for eResearch (host institute) and the School of Environment (co-located institute), I will develop and implement an engaging cross-platform location based game to crowdsource landscape perception and preference data. Prof. Dr. Mark Gahegan (hosting supervisor) has detailed knowledge of developing spatial applications (Adams et al., 2015; Whitehead & Gahegan, 2012), including for a younger audience (Fuhrmann et al., 2005) and has in-depth expertise in the discipline of geographic information science (Gahegan, 2020; Harris et al., 2017) as well as experiences in natural language processing (Tan et al., 2021). These are invaluable expertise for the development of the proposed LBG and the analyses of the contributions. Real-world data will be incorporated into procedurally building the game world and special attention will be paid to user privacy in compliance with national and international laws. The location based game will seamlessly integrate both simple (categorical lists) as well as more complex (landscape descriptions & photographs) tasks to increase immersion and resulting motivation. The application will be promoted throughout our academic networks to ensure a sufficiently large number of participants. Additional efforts will go towards promoting the application in the host country to capitalise on New Zealand’s unique topography as well as cultural diversity, with the Māori people, the European settlers as well as a vibrant community of immigrants. Motivating participants from various cultural and socio-demographic backgrounds will allow unprecedented inquiries into language, culture and landscape perceptions. 

Outcome: A globally accessible active crowdsourcing platform with intriguing gameplay to motivate users to generate rich landscape-relevant information. Key implementational decisions and resulting implications for scientific data generation as well as correlations between contributions and promotional efforts will be reported in scientific articles.

Exploring participants and landscapes

Meta-data will be collected on participants (e.g. demographics & home location) as well as on contributions (e.g. location & time) allowing in-depth analyses of user characteristics and shedding light on the targeted audience. The success of promotional efforts will be gauged and a set of recommendations based on descriptive statistics derived from the implemented application will be published to support future participatory data generation efforts. Further, the collected meta-data will be used to cluster individuals into socio-demographic and cultural groups as well as local and visiting populations to explore inter- and intra-group similarities and differences in contributions. The application will collect user-reported landscape classifications according to predefined categories (e.g. cultural ecosystem services) as well as natural language contributions (e.g. landscape descriptions) and images (e.g. representative landscape images). These contributions will allow detailed explorations of how landscapes and specific elements and dimensions therein are experienced, perceived and valued and build the foundation for a multitude of further research avenues. In addition, participants’ stated (e.g. contributions) and revealed (e.g. frequented locations) (Adamowicz et al., 1994) landscape preferences will be communicated through scientific publications as well as public events and stakeholder meetings. Prof. Mark Gahegan’s vast academic network will open the door to various beneficial collaborations with researchers at the University of Auckland (Dr. Katarzyna Sila-Nowicka and Dr. Michael Martin) as well as other academic institutes in the host country (Prof. Dr. Ben Adams, Prof. Dr. David O’Sullivan and Dr. Lars Brabyn) resulting in an excellent interdisciplinary team to guide the implementation and, in particular, to explore the generated contributions and the landscape relevant information they capture.

Outcome: A collection of user-generated in-situ landscape reports of predefined categories, multiple corpora of landscape-relevant natural language and a collection of landscape images. In addition, participants’ stated and revealed landscape preferences will be explored in interdisciplinary collaborations and communicated in scientific publications to guide future policy and decision making processes.

Exploring landscape perceptions within and between individuals and groups

To explore landscape perceptions and preferences in detail, contributed classification reports will be analysed in their spatial configuration and tied to individual landscape perceptions. Further, natural language contributions will be annotated and analysed in terms of the tangible and intangible landscape relevant dimensions they capture. Specifically, I will combine qualitative coding and computational approaches using machine learning and natural language processing under the supervision of Prof. Dr. Mark Gahegan (Tan et al., 2021) and in collaboration with relevant researchers to identify and extract experiential and perceptual dimensions. The analysed contributions will inform the development of (automated) landscape perception extraction from text as well as support policy and decision making processes. Identifying similarities and differences in how individuals or groups of individuals perceive certain landscapes is becoming increasingly important, especially in fields of tension between economic exploitation and preservation of natural and anthropogenic landscapes. This research aims to contribute to arising societal questions of how we interact with our environment and the consequences thereof, transferable to any area of interest (e.g. How is deforestation for timber production perceived in New Zealand? How can landscapes be repurposed for wind- or hydroelectricity generation in Switzerland?). 

Outcome: An in-depth report on significantly similar and different landscape experiences, values, preferences and perceptions depending on socio-demographic and cultural attributes. Published findings will shed light on salient landscape dimensions as a function of participants’ backgrounds and place of residence, leading to much needed insights to guide future landscape related inquiries and address societal questions.

Relevance and impact

Many projects struggle with recruiting participants, mostly due to financial and time restrictions or lack of incentive to participate. Exploring the potential of gamification can lead to novel insights into user motivation and retention, transferable to all areas of academic participatory data generation. Gamification also offers a novel approach adaptable to new projects as these could be integrated into the proposed LBG (cf. MMOS, a Swiss gamification initiative generating billions of contributions in highly popular video games). This is of particular importance seeing that many academic participatory data generation platforms are not maintained past the initial funding period. Further, an open dialogue between academics and laypeople will be encouraged and the generated knowledge will be made accessible. This project will also lend itself to exemplifying various concepts in a teaching capacity (e.g. geospatial data acquisition and biasses). The interdisciplinarity of the project implies new insights in various areas, ranging from geospatial particularities in gamification, over including underrepresented groups in landscape perception research, to addressing important societal questions about the future of our surroundings. Seeing the potential of gamification to attract a large number of participants, this research will explore how individuals perceive landscapes on a hitherto unprecedented scale, opening the door to a plethora of new research avenues. In particular, this research will add underrepresented public perceptions to landscape governance, thus having wide ranging implications for policy and decision making processes, especially when dealing with more intricate or controversial environmental questions where data from traditional sensors no longer suffice.