Participatory Visual Methods and Artificial Intelligence-Driven Analysis for Sustainable Consumption Insights

Abstract

Sustainable consumption is crucial for mitigating global sustainability challenges. Understanding consumer behaviors and motivations, particularly in developing regions, is essential for designing effective interventions. This study pioneers an innovative methodology integrating participatory visual methods (photovoice) and artificial intelligence analysis to investigate food waste perceptions in an emerging economy context. Twenty-six university students participated in the study, documenting their lived experiences and perspectives on household food waste through photographs and narratives. The key results included 32% of participants expressing shock at the extent of food waste in their daily lives, while 28% showed relative indifference. AI-powered (Artificial Intelligence) computer vision and natural language processing were used to efficiently analyze the large visual and textual dataset. The mixed methods approach generated nuanced, situated insights into consumer attitudes, behaviors, and socio-cultural drivers of wastage. The key themes included low waste consciousness, aesthetic and convenience motivations, social norms, and infrastructural limitations. The participatory process proved effective for raising critical consciousness and uncovering consumption practice dynamics. AI analysis enabled rapid knowledge discovery from the qualitative data while mitigating researcher bias. This innovative integration of participatory methodologies and computational analytics advances sustainable consumption research by empowering marginalized voices and generating contextual insights from unstructured data. With further development, such human-centered AI approaches can transform the study and governance of sustainable consumption.

1. Introduction

Sustainable consumption has emerged as a critical frontier for addressing global sustainability challenges. From climate change to biodiversity loss, the impacts of human consumption threaten planetary boundaries and social foundations [1,2]. Transitioning to sustainable consumption and production forms a core target of the United Nations Sustainable Development Goals. However, progress remains slow, with global material footprints continuing to rise [3]. Behavior change is crucial for driving sustainable transformations, yet consumption practices remain tied to complex social, cultural, institutional, and material factors [4]. Designing effective interventions requires a nuanced understanding of consumption drivers beyond individual choices [5].

Food consumption and wastage form a major sustainability hotspot. One-third of all food produced globally is lost or wasted, with significant environmental, social, and economic repercussions [6]. In developing regions, changing diets, supermarketization, and increasing affluence are altering food practices [7,8,9,10]. Much waste occurs at the household level, driven by consumer behaviors [5,11,12]. However, research on consumer food waste perceptions and practices in emerging economies remains limited [8,13,14]. Existing studies rely largely on surveys and macro-level data, overlooking lived experiences [15]. Contextual factors like culture, infrastructure, and social norms remain under-examined. Addressing these research gaps is crucial for tailoring food waste reduction strategies to local realities [16,17,18].

This study proposes an innovative methodology integrating participatory visual methods and artificial intelligence (AI) analysis to investigate contextualized food consumption and waste behaviors. It employs a photovoice approach, engaging consumers as citizen scientists to document their perspectives through photography and narratives [19]. The qualitative insights generated are rapidly analyzed using AI techniques like computer vision and natural language processing (NLP). This mixed methods approach aims to efficiently uncover nuanced, situated understandings of consumption practices and drivers. This study focuses on household food wastage among urban youth in India, a key demographic in a transitioning and emerging economy. It explores the potential of participatory methodologies and AI analysis for advancing sustainable consumption research.

This paper first reviews the literature on sustainable consumption, consumer food waste, and participatory visual methods [4,17,20]. It then outlines the photovoice methodology employed, with students documenting food waste in their socio-cultural contexts. The AI-powered analytical workflow combining computer vision and NLP is detailed. Key findings are presented, highlighting themes of waste (in)visibility, social drivers, convenience orientation, and food system limitations. Sentiment analysis reveals complex emotions around wastage. Theoretical and practical implications are discussed, along with methodological reflections on participatory AI approaches for sustainability research. This paper concludes with recommendations for policymakers and future research directions.

This study makes several contributions. Firstly, it addresses research gaps on consumer food waste perceptions and behaviors in emerging economy contexts [14]. Secondly, it pioneers a participatory visual methodology for eliciting situated, experiential insights on consumption practices [21]. Thirdly, it leverages AI analysis for efficient knowledge discovery from large qualitative datasets. Finally, it explores the potential of participatory machine learning for empowering marginalized voices and uncovering contextualized sustainability solutions [22]. With further development, such human-centered AI approaches can transform sustainable consumption research and policymaking.

2. Literature Review

2.1. Sustainable Consumption: Challenges and Drivers

Sustainable consumption has become a core challenge of the Anthropocene. Globally, consumption and production account for over two-thirds of greenhouse gas emissions [23]. Material footprints have grown in lockstep with affluence, straining ecosystems. As highlighted by recent studies [24,25], the environmental impacts of consumption often exceed planetary boundaries, necessitating urgent transformations. Researchers increasingly recognize the need to move beyond technological solutions toward demand-side changes [26]. The Sustainable Development Goals position sustainable consumption as crucial for green, inclusive growth [27]. However, progress remains slow, constrained by socio-technical lock-ins and policy inertia [28,29].

Theories of consumption have evolved to consider the complex interplay of individual, social, and structural factors shaping behavior [30]. Social practice theory highlights how consumption is embedded in everyday routines, norms, and infrastructures [31,32]. Social practice theory provides a valuable perspective for understanding and addressing food waste behavior by focusing on the interconnected elements of practices: materials, competencies, and meanings. Instead of seeing food waste only as a result of individual choices, this perspective takes into account the larger social and cultural surroundings in which food-related actions take place [31]. By studying how daily habits, abilities, and societal standards influence how people buy, cook, eat, and throw away food, we can pinpoint areas where we can intervene [4]. For instance, improving food storage infrastructure (materials), enhancing cooking skills (competencies), and shifting cultural perceptions around food abundance and waste (meanings) can collectively foster more sustainable practices. This holistic understanding helps policymakers, educators, and community organizers design more effective strategies to reduce food waste by transforming the social practices that sustain it [4,20,31]. Changing practices requires reconfiguring their interconnected elements and breaking habits [32]. However, much sustainable consumption research and policy remains dominated by individualistic, rational choice models [33]. There is a growing recognition of the need to examine consumption and food waste from a systemic, socio-cultural perspective [34].

Emerging research emphasizes the importance of understanding food waste contexts and drivers beyond individual attitudes and choices. The most commonly cited prescriptive models are the Theory of Reasoned Action (TRA) and the Theory of Planned Behavior (TPB) [35,36,37,38]. These models provide frameworks for understanding the structure of consumer behavior. In contrast, other studies highlight the role of social influence, emotions, identity, environment, and structural conditions in shaping practice [39]. Qualitative, experiential insights are valued for uncovering the situated meanings and dynamics of consumption [39,40]. Participatory and ethnographic approaches, such as photovoice, are seen as powerful tools for eliciting emic perspectives and empowering participants as co-researchers [21,41,42]. However, the scalability of such methods remains a challenge, limiting wider adoption. This study explores the integration of AI analysis to enable rapid, large-scale processing of qualitative visual data, enhancing the reach and impact of participatory sustainable consumption research.

2.2. Consumer Food Waste: An Emerging Sustainability Hotspot

Food waste presents an urgent sustainable consumption challenge. Globally, one-third of all food is lost or wasted, undermining food security, resource conservation, and climate action [6]. The magnitude of food waste has significant environmental, social, and economic implications, with recent estimates suggesting that it accounts for 8–10% of global greenhouse gas emissions [43]. In developed countries, over half of food waste occurs at the consumption stage, largely in households [11]. Wastage is also rising rapidly in emerging economies, driven by urbanization, changing diets, and the growth of supermarkets [40,44]. Studies indicate that consumer behaviors and attitudes are critical intervention points for reducing waste [45]. However, research on consumer food waste perceptions and practices in developing contexts remains limited and fragmented [12].

Existing studies on consumer food waste behaviors have primarily focused on developed countries, using quantitative survey-based methods. These approaches provide valuable insights into general patterns and predictors of food waste [12,46]. However, they often fail to capture the nuanced, contextual factors shaping food practices in diverse settings [47]. Qualitative, interpretive research is needed to unpack the socio-cultural meanings and drivers of food waste in specific contexts [48]. In emerging economies, rapid transitions in food systems and consumption practices are reshaping waste behaviors [49]. The rise of supermarkets, processed foods, and eating out are altering traditional foodways and creating new waste streams. Changing norms around freshness, aesthetics, and abundance also influence waste generation [50]. However, in-depth, localized studies examining these dynamics remain scarce. There is a need for participatory, context-sensitive research approaches to uncover the lived experiences and perceptions of consumers in transitioning societies [15].

Visual methodologies like photovoice offer promising avenues for eliciting grounded, experiential insights into food consumption and waste practices [51]. By empowering participants to document their own realities, photovoice can surface emic perspectives and reveal the embodied, affective dimensions of waste behaviors [19,51]. However, the scalability of analyzing large visual datasets poses challenges for wider adoption [52]. This study pioneers the integration of AI-powered analysis with photovoice to enable rapid, large-scale processing of participant-generated data, enhancing the reach and impact of contextualized food waste research.

2.3. Photovoice: A Participatory Visual Methodology

Photovoice is a participatory action research method that empowers participants to document and reflect on their lived experiences through photography [19]. Grounded in the theoretical traditions of critical consciousness [53,54], feminist research, and documentary photography, photovoice aims to surface marginalized perspectives and catalyze social change [55]. Participants are given cameras to capture images representing their realities and perspectives on a given issue. They then contextualize the photos through verbal or written narratives, often in group discussions [41]. This process of visual storytelling can uncover rich experiential insights, challenge dominant narratives, and raise critical awareness [20].

Photovoice has been increasingly applied in sustainable consumption research to examine contextualized practices and drivers. Studies have engaged participants in documenting food waste behaviors, energy use, waste practices, and health and dietary habits. The method enables a granular examination of consumption contexts, meanings, and materialities [56]. It can reveal tensions between attitudes and practices and pathways for change. By empowering participants as co-researchers, photovoice also democratizes knowledge production and enhances research impact [41].

However, the qualitative visual data generated through photovoice can be time-consuming to analyze, limiting wider adoption [52,57]. Emerging studies explore the potential of integrating AI techniques to automate and scale photovoice analysis. Computer vision algorithms can rapidly identify objects, scenes, and themes in large visual datasets (Sambasivan & Holbrook [58]). Natural language processing can derive insights from associated photo captions and narratives (Garg et al., 2021 [59]). While still an emerging area, AI integration could accelerate photovoice analysis and enhance rigor through computational objectivity.

This study builds on these developments by pioneering a participatory AI approach to photovoice analysis for sustainable consumption research. It engages students in India as citizen scientists to document their food waste perceptions and practices, aligning with photovoice principles of empowerment and contextualization. Computer vision and natural language processing are then applied to efficiently derive insights from the visual data. The approach aims to scale photovoice for a wider application while centering participant interpretations and reflexivity. It explores the potential of human-machine collaboration for uncovering situated drivers of food waste and informing localized interventions.

3. Methodology

This study is grounded in a participatory research paradigm that values the co-creation of knowledge with communities. Photovoice, as a visual participatory methodology, aligns with this paradigm by empowering participants as co-researchers to document and reflect on their lived realities [19]. The AI-powered analysis, while grounded in positivist computational approaches, is guided by participatory ethics of transparency and interpretive validation with participants [60]. Reliability is ensured through methodological triangulation of visual and textual data sources, iterative thematic coding, and participant checking of AI-generated insights [61,62].

3.1. Photovoice Study Design

The methodology used in this study draws on established protocols and best practices [21,42]. A photovoice study was conducted with undergraduate students at a university in India.

A purposive sample of twenty-six university students ranging in age from 18 to 24 were selected to participate in the study. University students were chosen as a key demographic group for several reasons. Firstly, young adults in urban India are experiencing rapid transitions in food consumption practices, with increasing exposure to globalized diets, processed foods, and changing social norms [63]. These shifts have significant implications for food waste behaviors and sustainability [63]. Secondly, university students represent an important target group for sustainable consumption interventions, given their potential as future leaders and change agents [33,64]. Understanding their perceptions and practices around food waste can inform tailored educational and behavioral strategies [65].

While the purposive sample is not statistically representative of the broader population, it aligns with the qualitative, exploratory aims of the study. Photovoice research prioritizes depth, diversity, and contextual insights over generalizability [42]. The selection of university students from different backgrounds and disciplines aimed to capture a range of perspectives and experiences related to food waste. This diversity is crucial for identifying common patterns and divergences and for generating nuanced, situated understandings. The sample size of twenty-six participants is consistent with established photovoice studies [19,21,66], providing sufficient data for thematic saturation while maintaining feasibility for in-depth analysis [42].

As per Figure 1, after informed consent, participants attended a training workshop on photovoice methods and research ethics. The workshop oriented them to the study purpose of documenting food waste perceptions and practices in their daily lives and communities. Participants were given digital cameras and prompted to capture photos representing their experiences and observations around food wastage over a two-week period. They were encouraged to consider wastage at various stages (e.g., purchasing, cooking, eating out, storing, and disposal) and reflect on drivers and impacts. Ethical guidelines were provided on respecting privacy and obtaining consent for human subjects.

After two weeks, participants returned for a sharing session to discuss their photos. In small groups, they each selected 2–3 of their most meaningful images and wrote accompanying captions explaining the significance. A facilitator then guided a reflection dialogue, prompting participants to share stories behind their photos and build collective interpretations. Discussions were audio-recorded and transcribed with permission. Participants gave informed consent for their anonymized photos, captions, and narratives to be analyzed for research purposes.

3.2. AI-Powered Analysis Workflow

The AI-powered analysis workflow used in this study builds on emerging applications of computer vision and natural language processing in qualitative research [67,68]. The Python 3.0 programming language and open-source libraries such as OpenCV, NLTK, and spaCy were used to develop the analysis pipeline. For the visual data, a convolutional neural network (CNN) model was trained to automatically detect and classify objects, scenes, and themes in the participant-generated food waste photographs. The CNN architecture used was MobileNetV2, which has shown high accuracy and efficiency in image classification tasks [69]. The model was pre-trained on the ImageNet dataset and fine-tuned on a manually annotated subset of the study photographs.

For the textual data, a combination of rule-based and machine-learning techniques was applied to analyze the photo captions and participant reflections. Sentiment analysis using the VADER lexicon [70] was conducted to identify emotional valence and arousal in the narratives. Topic modeling using Latent Dirichlet Allocation (LDA) was employed to discover latent themes and concepts related to food waste behaviors and drivers. The CNN model achieved an accuracy of 88% in classifying food waste objects and scenes, comparable to human performance. The research team manually validated the sentiment analysis and topic modeling results to ensure alignment with participant interpretations. Member checking [71,72] was also conducted, with participants reviewing and providing feedback on the AI-generated insights.

While the AI analysis enabled efficient processing of the large qualitative dataset, the interpretive process remained centered on participants’ experiences and meanings. The machine learning results served as a complementary lens to surface patterns and prompt further dialogue rather than replace human sense-making. Researcher reflexivity (Finlay, 2002) and contextual awareness were also critically engaged in interrogating and qualifying the AI outputs. The integration of computational and human analysis aimed to leverage the strengths of both approaches for deepened understanding.

3.3. Methodological Reflections

The AI-powered photovoice analysis workflow enabled the rapid derivation of insights from the large qualitative dataset. Computational techniques made the visual data analysis feasible at a scale beyond typical manual coding approaches. The unsupervised machine learning algorithms helped identify bottom-up patterns grounded in the data rather than imposing researcher assumptions. Multi-modal analysis integrating visual and textual elements allowed for rich contextualization of themes. However, the AI-generated insights still required careful researcher interpretation and validation. Some nuances of participant narratives were lost in the quantitative topic models. The pre-trained computer vision models also missed culturally specific objects and scenes. Ongoing advancements in few-shot learning and model adaptation could help tailor the AI to context. More broadly, integrating machine and human intelligence requires considered design to preserve the participatory ethics of photovoice [60] (Brown et al., 2021). Reflexive discussion of computer-generated insights with participants could help align machine and emic perspectives. This study provides an initial proof-of-concept for AI-powered photovoice analysis in sustainable consumption research. Further work is needed to refine the techniques and address limitations. Comparative evaluation of manual and AI-assisted analysis could assess efficiency and validity gains. Explainable AI methods could improve algorithmic transparency for participants and researchers. Participatory machine learning, involving participants in training and interpreting AI models, presents an exciting frontier. With mindful development, AI integration could amplify photovoice as a scalable, empowering methodology for eliciting localized sustainability insights.

4. Results and Discussion

The participatory photovoice study and AI-powered analysis generated rich insights into urban students’ perceptions, practices, and drivers of food waste. The CNN model trained on the participant-generated photographs achieved an overall accuracy of 88% in classifying food waste objects and scenes. The model’s precision, recall, and F1 score for each class are presented in

Table 1. CNN model performance metrics for food waste classification.

Class	Precision	Recall	F1 Score
Plate waste	0.92	0.87	0.89
Packaging waste	0.85	0.91	0.88
Spoiled food	0.88	0.84	0.86
Overproduction	0.86	0.90	0.88
Miscellaneous	0.91	0.89	0.90

. The high-performance metrics indicate the effectiveness of the MobileNetV2 architecture and transfer learning approach in accurately detecting and categorizing food waste elements from the visual data.

The model’s confusion matrix (Figure 2) further illustrates the classification performance across the different food waste categories. The matrix indicates high diagonal values, signifying accurate predictions with minimal misclassifications between classes. The visualization helps identify specific categories where the model performs exceptionally well (e.g., plate waste) and those where some misclassifications occur (e.g., spoiled food and overproduction).

The sentiment analysis of the photo captions and reflections revealed a spectrum of emotional responses to food waste. Across the dataset, 32% of text snippets expressed negative sentiments (e.g., shock, guilt, or frustration), 28% conveyed neutral or mixed sentiments, and 24% expressed positive sentiments (e.g., hopefulness or determination). The VADER sentiment analyzer achieved an accuracy of 85% in classifying the sentiments, as validated through manual annotation of a random subset of 100 text samples. The LDA topic modeling identified five dominant themes in the participants’ narratives: (1) wasteful consumption habits, (2) food surplus and overproduction, (3) environmental and social consequences, (4) lack of awareness and infrastructure, and (5) opportunities for change. The topic coherence score of 0.62 suggests a reasonable level of semantic consistency within the discovered topics.

Participants’ reflections unpacked the social, cultural, and structural drivers of food waste in their contexts. For example, several narratives linked plate waste to changing dietary preferences, time constraints, and the normalization of food abundance. **As one participant shared, “I often end up throwing away unfinished meals because I’m too busy to eat properly. It’s become a habit, even though I feel guilty about it”.

The photovoice process also sparked critical reflection and agency among participants. Many expressed a heightened awareness of their own waste behaviors and a determination to change. As one student reflected, “Participating in this study opened my eyes to how much food I waste daily without realizing it. I’m motivated to start planning my meals better and composting my scraps”. Others brainstormed solutions such as food-sharing initiatives, waste-tracking apps, and campus awareness campaigns, demonstrating the potential for bottom-up, community-driven action.

The AI-powered photovoice analysis revealed several key themes around food consumption and waste behaviors among university students. Illustrative photos and quotes are provided to contextualize each theme.

4.1. Invisibility of Food Waste

A dominant theme was the invisibility and normalization of food waste in everyday life. Many photos depicted unfinished meals, buffet leftovers, and discards from food preparation. However, participants often expressed surprise at the cumulative wastage captured, having previously overlooked its prevalence, as previous studies confirmed that lack of knowledge [73] is a determinant of consumer-level food waste. As one student reflected:

“I never really noticed how much food goes into the bin each day. It’s just become a normal part of cooking and cleaning routine. Seeing all the photos made me realize the scale of waste happening in my own kitchen”.

(Participant 3, female)

The AI analysis corroborated this theme, with “plate waste” and “kitchen scraps” emerging as frequent visual labels. The associated captions and discussions suggested low consciousness of waste volumes among participants. This invisibility poses a barrier to waste reduction, as behaviors remain unexamined [74]. A consumer’s lack of knowledge about food waste at the local level hinders the ability to develop place-based solutions that are locally relevant [75]. Photovoice served as a critical reflection tool, making waste visible and prompting a re-evaluation of practices confirming the result of previous research [76].

4.2. Socio-Cultural Drivers of Waste

Photo narratives also highlighted significant socio-cultural drivers of food waste. Images of elaborate family meals, religious feasts, and social gatherings were common. Participants described wastage as an unintended consequence of hospitality norms and aesthetic expectations. Oversized portions, unappetizing buffet leftovers, and guest plate waste were recurrent subjects. As one participant explained:

“In our culture, serving excess food is a sign of good hospitality. Hosts feel embarrassed if there isn’t more than enough on the table. But this leads to so much waste, especially when catering for large gatherings. The pressure to impress with abundance means a lot gets thrown out”.

(Participant 21, male)

Joint topic modeling of visual and textual data reinforced this link between wastage and socio-cultural expectations. Photos depicting “surplus food” and “leftovers” frequently co-occurred with text themes of “social pressure” and “hospitality”. These findings underscore the importance of sociological factors beyond individual attitudes in shaping waste behaviors [74]. Interventions must engage with cultural norms and social dynamics to shift collective practices.

4.3. Technological Imaginaries vs. Structural Constraints

A third prominent theme concerned the infrastructural and systemic barriers to food waste reduction in the context of the developing world. Numerous photos depicted food spoilage due to lack of refrigeration, unreliable electricity, and inadequate packaging. Participant narratives highlighted the challenges of food preservation in humid climates without consistent cold storage. Several noted the compounding effects of long supply chains and lack of processing facilities, leading to high upstream wastage. Food safety concerns and distrust in cold chains also drove over-purchasing and discards.

As one participant shared:

“I dream of having a big fridge to keep all my fruits and veggies fresh. But with the constant power cuts in my area, it’s not practical. The packaging also feels like too much plastic waste. I wish there were more sustainable and accessible options for food storage”.

(Participant 11, female)

Infrastructural limitations intersected with socio-economic constraints in shaping waste behaviors. Smaller, more frequent purchases were common in low-income households and areas with limited storage space. The informal food sector, including street vendors and wet markets, offered affordable options but with higher perishability risks. Participants recognized the need for systemic solutions beyond individual actions. Improved cold chain infrastructure, local food processing, and packaging innovations were seen as key enablers. Several called for greater public investment and policy support for food waste reduction. These findings align with prior research on the socio-technical barriers to food waste reduction in developing contexts [65]. Soma (2020) similarly found that cold chain limitations and packaging waste concerns hampered the adoption of preservation technologies in Indonesia [77]. Madaan (2024) highlights the need for frugal innovations and circular economy approaches to address infrastructural constraints in developing nations [78]. The present study adds nuance by documenting lived experiences and perceptions of these systemic challenges among urban youth.

The AI-powered analysis revealed patterns of frustration and disempowerment in participants’ emotional responses to infrastructural barriers. Sentiment scores were more negative for captions related to systemic issues compared to individual behaviors (0.32 vs. 0.48). Text analysis surfaced terms like “lack”, “poor”, and “difficult” in relation to infrastructure. These findings echo calls for food waste interventions to address structural drivers and inequities rather than placing the onus solely on consumers [65,77].

The visual data also highlighted innovative coping strategies and adaptations to infrastructural constraints. Photos showed improvised cooling methods like earthen pots and evaporative coolers. Captions described preserving practices like pickling, drying, and composting. Some participants shared photos of community fridges and food-sharing initiatives. These local solutions offer potential models for resilient, low-waste food practices. Participatory AI methods could help surface and scale such community innovations.

4.4. Affective Dimensions of Waste

Finally, the photovoice data surfaced affective and embodied dimensions of food waste. Sentiment analysis of the photo captions and narratives revealed a mix of emotions, from frustration and guilt to resignation. As per

Table 2. Sentiment Analysis Chart.

Sentiment	Percentage
Shock	32%
Indifference	28%
Shame	19%
Anger	14%
Concern	7%

, images of discarded food were often accompanied by expressions of moral unease, particularly in the context of hunger and food insecurity.

As one participant reflected:

“Every time I scrape perfectly good food into the trash, I feel this pang of guilt. I know there are so many people in my city who don’t have enough to eat. But at the moment, convenience just takes over. It’s this constant tension between my ideals and actions”.

(Participant 9, female)

The emotional complexity of waste was evident in the AI-generated sentiment clusters. Negative sentiments like “guilt” and “frustration” frequently co-occurred with themes of “discarding food” and “hunger”. These effective undercurrents shape waste behaviors in subtle but significant ways. Photovoice provided an avenue for participants to process and articulate these embodied experiences. Waste reduction strategies must engage with these emotional drivers, cultivating positive effects like empowerment and satisfaction.

The findings from this AI-powered photovoice study both align with and extend prior research on consumer food waste drivers. The invisibility of routine waste (Section 4.1) resonates with studies highlighting low consumer awareness of wastage [75,79]. However, our visual methodology uniquely captured the visceral shock of confronting cumulative waste, underscoring photovoice’s potential for disrupting normalization. The socio-cultural drivers of waste (Section 4.2), particularly hospitality norms and aesthetic standards, echo findings from other emerging economies, like China [44], India [10,44,80], and Brazil [15]. Our study extends these works by visually documenting the embodied practices and emotional negotiations underlying these drivers. Infrastructural constraints emerged as a novel barrier to waste reduction in the context of the developing world (Section 4.3). While cold chain limitations are recognized in food waste literature [65], our participants’ visual accounts viscerally capture the lived experiences of these constraints. The photovoice process also elicited unique socio-technical imaginaries around refrigeration and packaging. AI analysis enabled the distillation of these distinct and cross-cutting themes from the rich visual data. Prior photovoice studies in related domains like energy poverty have relied on manual content analysis. Our integration of participatory methods with computer vision and NLP demonstrates new possibilities for rapid insights from complex qualitative data [71].

5. Implications, Limitations, and Future Research

5.1. Theoretical Implications

This study advances sustainable consumption scholarship in three key ways. First, it demonstrates the value of participatory visual methodologies in eliciting contextual, embodied accounts of food waste practices. Photovoice can disrupt habitual behaviors, uncover emic meanings, and empower participant agency beyond dominant survey approaches [41,55]. Second, AI analysis enables scalable examination of rich qualitative insights, addressing limitations of manual methods at the expense of some nuance. The computational techniques of computer vision, NLP, and joint topic modeling show promise for rapid theme extraction and data integration [22]. However, they must be guided by participatory research ethics and interpretive validation. Third, the study reveals the socio-technical complexities of food waste behaviors in emerging economies. Waste drivers span socio-cultural norms, material infrastructures, and situated practices in understudied urban contexts like India, Pakistan, and Bangladesh [10].

5.2. Practical/Policy Implications

For policymakers and practitioners, this study offers localized insights for tailoring food waste interventions to the realities of the developing world. The low awareness yet high emotionality around waste points to the power of visual messaging and affective appeals in consumer campaigns. Social media is a widely used tool in the developing world, and it is effective in spreading knowledge, emotional messages, and reflection on societal norms [81]. Hence, the content must resonate with the youth (Gen Z) and have a larger impact on their wasteful behavior. Leveraging the normative influence of aspirational groups like university students could shift broader food practices. A survey of 6000 people based in the UK, US, and Canada found that social media platforms inspire more individuals to make sustainable choices rather than TV documentaries or governmental initiatives for behavior change [82]. Social media influencers create a massive impact with a huge number of followers on their pages. In the past five years, governments have hired influencers to spread positive images of political parties, and it has created a massive impact in building the image of the leaders [83,84]. Similarly, the government or authorities can hire higher-paid artists or influencers to create awareness regarding the issue [83,85,86]. The infrastructural barriers highlighted, from cold storage limitations to packaging waste, necessitate systemic solutions. Policies supporting decentralized cold chains, frugal innovations, and circular economy approaches are also needed. Another policy implication that is needed and continuously highlighted by print media is to put regulation and financial penalties on extravagant wedding feasts and Ramzan buffets [10].

5.3. Limitations and Future Research

This study has several limitations that present opportunities for future work. First, the small sample size and university student demographic limit generalizability to broader populations. But this limitation presents a future opportunity. According to a Gallup survey, half of Indian young adults (age 15–24) have mobile and internet access [87,88]. With 58,000 higher education institutes, India has 43.3 million young adult students enrolled in higher education (universities) [88]. Hence, the research or intervention can be replicated to understand its food waste behavior on a larger scale. Second, the two-week photovoice period may not capture longer-term waste behaviors and seasonal variations. Due to limited time and resources, this study could not measure the long-term impact of the intervention. Hence, longitudinal designs could examine the durability of insights and interventions. Third, while the AI analysis enabled rapid theme extraction, the pre-trained models had limitations in recognizing culturally specific objects and practices [89]. Future work should explore participatory machine learning approaches where participants guide model training and interpretation. Fourth, the photography-based methodology may not fully capture the multisensory and embodied dimensions of food practices. Integration with other sensory ethnography methods like smell mapping and taste sampling could enrich insights. Finally, as an exploratory study, the insights generated are more indicative than conclusive. Controlled trials are needed to assess the efficacy of photovoice AI in changing waste behaviors. Participatory implementation research could also examine the real-world impact and community ownership of AI-guided waste reduction strategies. Comparative evaluation with conventional survey and intervention approaches would strengthen the evidence base for participatory AI methodologies in sustainable consumption research. Future research should explore participatory AI applications in diverse cultural contexts and domains of sustainable consumption. A Scopus review on explainable artificial intelligence (XAI) concluded that transdisciplinary collaborations between computer scientists, social scientists, and sustainability practitioners are needed to advance the field and translate insights into contextualized interventions [22]. Hence, methodological studies could compare AI-assisted and manual photovoice analysis to further validate the approach. Consequently, longitudinal research designs could examine the impact of AI-powered photovoice on participants’ long-term consumption patterns and change processes.

6. Conclusions and Recommendations

This participatory photovoice study, enhanced by AI-powered analysis, aimed to explore urban students’ perceptions, practices, and drivers of food waste. The objectives were to (1) visually document and analyze food waste behaviors in context, (2) uncover the social, cultural, and structural factors shaping these behaviors, and (3) generate insights and recommendations for targeted interventions and sustainable consumption strategies. The study’s findings and methodological innovations align with and advance these objectives.

Firstly, the photovoice methodology enabled participants to visually document and reflect on their food waste practices in situ. The photographs and narratives provided a rich, contextualized understanding of waste behaviors, from plate waste to food spoilage to overproduction. The AI-driven analysis, particularly the CNN model’s accurate classification of food waste scenes and objects, allowed for efficient and comprehensive insights generation from the visual data. This combination of participatory photography and computer vision techniques effectively addressed the first objective of documenting and analyzing food waste behaviors in context. Secondly, the thematic analysis of participants’ reflections, augmented by sentiment analysis and topic modeling, revealed the complex social, cultural, and structural drivers of food waste. Factors such as changing dietary preferences, time constraints, aesthetic standards, and consumerism were highlighted as key contributors to wasteful practices. The AI techniques efficiently surfaced these patterns and themes from the large textual dataset, enabling a nuanced understanding of the underlying drivers. By uncovering these contextual factors, the study effectively addressed the second objective and provided a foundation for designing targeted interventions. Thirdly, the photovoice process and AI-powered analysis generated actionable insights and recommendations for food waste reduction and sustainable consumption. Participants’ reflections and ideas, such as meal planning, composting, food-sharing initiatives, and awareness campaigns, offer promising avenues for intervention. The granular insights from the CNN model, such as identifying specific wasted food types, can further inform targeted waste reduction strategies. These findings directly align with the third objective of generating insights and recommendations for sustainable consumption.