Audience-Oriented Information Intervention Approach to Food Waste Behavior: An Application in Chinese University Canteens

Abstract

Information intervention is often used to reduce food waste in restaurants and canteens. Existing studies have overlooked the target audience’s psychological and behavioral characteristics during message design, resulting in erratic intervention outcomes. This study proposes an audience-oriented information intervention approach that integrates goal systems theory and information feedback paradigms. The cornerstone underlying this approach is the elaboration of release-type and feedback-type intervention messages. Our approach offers a procedure to determine the optimal message-based intervention program. This study empirically applied the approach to intervene in food waste behavior among 1141 Chinese university students in canteens. A between-subjects design quasi experiment was conducted to explore the effect of information intervention, and the results showed that the downward comparative feedback-type message at the ordering stage performed the best. Findings offer practical guidance for reducing food waste in various consumption scenarios and useful inputs for encouraging green consumption behavior.

1. Introduction

The environmental, social, and economic repercussions of food waste are far-reaching [1,2,3]. The Food and Agriculture Organization of the United Nations (FAO) estimates that 1.3 billion tons, or one-third, of the world’s food is lost or wasted from its initial production to final consumption—this amount could feed one-ninth of the world’s undernourished populace [4,5,6]. Sustainable Development Goal 12.3 from the United Nations [7] emphasizes the importance of curbing food waste. It sets a goal to halve per capita global food waste at the retail and consumption stages by 2030, while also minimizing food loss in the production and supply sectors [8]. Based on current trends, if the global population reaches 9.8 billion by 2050, sustaining our current lifestyle would require natural resources three times the Earth’s capacity [7]. In order to feed the rapidly growing population with limited resources, it is imperative to take collective action to reduce food waste.

Information intervention is a cost-effective measure for reducing food waste [9]. Previous studies examining the impact of information intervention on food waste have often considered characteristics of the message content like communication modality (e.g., written vs. verbal) [9], message frame [10], how and why messages [11], message concern like message focus (on close versus distant social actors) [12], and intervention timing like presentation order (before meals vs. during meals) [9]. We focus on the message content and presentation order.

This study used two types of intervention messages, namely the release-type intervention message and the feedback-type intervention message. Previous studies have not reached a consensus on the effectiveness of these two message types in reducing food waste, but most have highlighted the positive impact of these two types of intervention message [1,13,14]. On the other hand, some studies have reported opposite findings [15].

The release-type message serves straightforward advocacy information about reducing food waste. This type of message typically features straightforward content and does not necessitate readers’ engagement with deep logical thinking. Examples include “All Taste No Waste”, “Eat What You Take. Don’t Waste Food” [13], “Ugly Fruits and Veggies” [16], “Clean Your Plate” [15], and “Love your food, down to the last bite” [17]. The content of these messages is largely derived from previous literature and real-world applied labels, and in some articles the rationale for the messages may not be clearly elucidated. These message contents were probably not tailored to the specific audience and might be a little lacking in pertinence [13,15,17]. Previous research indicated that tailored intervention messages receive more attention than non-tailored ones, suggesting that individuals might engage more deeply with tailored messages [18]. The aim of information intervention is persuasion [19]. As the communication theory highlights, information flows in two directions, and the primary focus of communication is the subjective interpretation of the message. A message holds meaning only when interpreted, and the creation of meaning emerges from the interplay between messages, participants (both senders and receivers), and contextual factors [20,21]. Crucially, both senders and receivers are equally active participants in the communication process [20]. Previous studies may have often assumed a generic and abstract audience when designing intervention message content, thereby overlooking the specific needs of the target audience [15,17]. In short, current intervention messages tend to function as a unidirectional dissemination of information. Audience orientation should offer valuable insights, enabling individuals to better comprehend and address food waste [22]. Feedback-type intervention messages furnish comprehensive data regarding behavior, encompassing daily food waste statistics, the number of malnourished individuals the wasted food could sustain, and individual waste amounts [13]. While this type of message is audience-oriented, it tends to focus solely on the behavioral outcomes of the immediate target audience [23], ignoring the antecedents and processes of food waste behavior, making the message relatively narrow in scope.

2. An Audience-Oriented Approach to Information Intervention

Passalacqua [24] emphasized the importance of correspondence between information content and audience demand in changing audience behavior. Many psychologists and other social scientists have mentioned the purpose of human behavior [25]. However, when exploring a message’s persuasion effect, the purpose of human behavior is largely ignored [26]. This study focused on the target audience themselves and the entire process of the target behavior (including before, during, and after the implementation of the behavior), considering the audience’s behavior goals and behavior results. This article used the goal systems theory to gain insight into the target audience’s cognition and motivation regarding the target behavior, and designed release-type intervention messages. Using self-feedback and comparative feedback paradigms, this paper designed feedback-type intervention messages that reflect behavioral performance. At the same time, relying on the contextual behavioral process, this study also considered influencing factors such as the form, presentation order, and display position of different intervention messages. Finally, through controlled experiments, the optimal information intervention program was identified. The audience-oriented food waste behavior information intervention approach proposed in this study specifically includes the following three parts.

2.1. Release-Type Intervention Messages Based on the Goal Systems Theory

Human actions are largely driven by goals [27,28]. According to the goal systems theory, the number of goals linked to a given means defines the multifinality set, and the size of the multifinality set may partially affect the perceived value of the given means [28,29]. The notion of subjective utility suggests that the means to be chosen might often be the one that promises to deliver the highest value or the utmost “bang” for the psychological “buck” [29]. Often, this could be a means characterized by maximal multifinality [28]. We can reasonably infer that cognitively activating the subgoals of a given behavior may expand its multifinality, thereby enhancing its perceived value and promoting individuals to engage in it.

Human cognition is influenced by contextual framing effects, and so are the goal systems [28]. Context can establish, modify, or eliminate cognitive associations between elements of the goal systems [28,30]. Goals and means can activate each other [31]. In this process, the strength of the cognitive association between the goal and the means not only holds significance but is also directly proportional to the activation strength [28]. The strength of the goal–means cognitive association is influenced by repeated pairing. This repeated pairing can be further strengthened by authoritative expert claims [28,30]. We believe it can also be realized through contextual messages, an external source of learning. Therefore, this study posits that the use of intervention messages focused on potential subgoals can activate and strengthen the goals–means cognitive association in food waste reduction. For example, messages may promote ordering wisely by expanding the multifinality set and heightening perceived value, thereby increasing the likelihood of adopting particular behavior to curtail food waste.

Higgins [32] categorized goals as either promotion goals or prevention goals. Achieving promotion goals leads to feelings of happiness and pride, without eliciting feelings of sadness or depression. Achieving prevention goals results in feelings of calmness and relaxation, without inducing feelings of tension or anxiety. Meanwhile, message frames can affect the receivers’ decision-making [33]. Different goals necessitate the use of corresponding message frames to convey the content, thereby achieving a regulatory fit [34,35]. Specifically, presenting promotion (prevention) goals using gain (loss) frames can enhance message persuasiveness, bolster goal-striving motivation, and subsequently influence the likelihood of adopting specific means to reduce food waste. Therefore, this study holds that in the process of designing release-type intervention messages, the message frames of the goals should be carefully considered according to the situation and target audience.

2.2. Feedback-Type Intervention Messages

These messages target the past behavior of the individuals or groups being intervened upon. In this study, we categorized feedback into three types: self-feedback, upward comparison feedback, and downward comparison feedback. Self-feedback provides the audience with information about their past behavioral outcomes [23,36]. Comparative feedback contrasts an individual’s behavior with that of relevant peers. In societal contexts, individuals and groups often compare themselves with others who have similar target characteristics [37,38]. Upward comparison involves contrasting oneself with individuals or groups considered superior [39], while downward comparison entails contrasting oneself with those considered lesser or inferior [40]. This study believes that different types of feedback messages are crucial elements in shaping the content directed toward the audience.

2.3. Overview of the Audience-Oriented Approach to Information Intervention

Generally, effective communication with consumers to achieve successful intervention outcomes hinges on several key factors. The content design should specifically target the audience, consider the presentation method, and empirically identify the most effective messages. Our audience-oriented information intervention approach comprises five stages, as illustrated in Figure 1.

The first stage involves identifying the target audience based on the actual context. It recognizes and defines each behavioral pathway, culminating in the ultimate outcomes, thereby laying down a foundational contextual framework for subsequent intervention messages.

The second stage, guided by the goal systems theory, constructs a contextual goal system for intervention based on the goals–means cognitive association strength by employing in-depth interviews and quantitative research methods. Concurrently, the release-type messages are formulated by considering the alignment between goal types and message frames.

The third stage involves formulating the content of the feedback-type messages. It examines the actual behavioral outcomes in both the intervention and comparison targets, using methods such as on-site observations, self-reports, and secondary data analysis.

The fourth stage focuses on preparing pending intervention messages. Factors considered include font type [41], background color [42], message position [41], and semantic fluency [43].

The fifth stage ensures experimental ecological validity by controlling various contextual interference variables. It conducts randomized controlled trials to find the optimal intervention messages and outlines specific implementation strategies.

3. Application: A Study on Food Waste Behavior Intervention in Chinese University Canteens

3.1. Selection of Target Audience

Food waste can occur in various contexts, such as hotels [44], restaurants [12], supermarkets [45,46], households [47,48], and student canteens [2,49]. Food waste from university canteens has especially attracted great attention from academia and the public in both developed and developing countries [50,51]. Food waste in canteens or restaurants is commonly defined as the amount of food left on a consumer’s plate after a meal, also known as “plate waste” (PW) [17,50].

Food waste in university canteens is serious [1,50,52], but university students seldom seem to be aware of their PW [17]. This perception starkly contrasts with the observations of dining staff, who note that PW is a significant issue [17]. University campuses, given their large populations, can serve as microcosms of urban areas [50]. Insights gained from tackling PW within university settings could have implications for broader communities [53]. Therefore, it is imperative to assess the current state of food waste among university students and identify effective strategies for reduction.

As a comprehensive university under the Ministry of Education of China, Jilin University is one of the country’s largest universities with six campuses. It serves 74,022 students from across the nation and features 22 canteens, offering a diverse range of dishes to cater to various student dietary preferences [54]. Jilin University was therefore chosen as a representative sample to explore the effect of information intervention on food waste.

3.2. Identification and Determination of Target Behavior

The stages of food waste generated in university canteens include kitchen preparation, consumption, and the doggie bag stage (see Figure 2) [44,55,56]. The consumption stage can be divided into the ordering stage and the dining stage [9]. This study focuses solely on food waste during the consumption stage, omitting the kitchen preparation and doggie bag stages.

Research shows that ordering food is one aspect of food waste at the consumption stage [57]. Influenced by factors such as time pressure [15,17], economic aspects [53], moral norms [57], and “Mianzi” (Face) culture [58], excessive ordering is identified as a significant antecedent leading to food waste. Additionally, due to factors such as taste [17] and dish size [17], the inability to finish the food at the dining stage also serves as a significant antecedent to food waste. Therefore, this study focuses on the ordering and dining behaviors, promoting ordering wisely and advocating for the “Clean Your Plate Campaign” to reduce and prevent food waste.

3.3. The Release-Type Intervention Messages

First, we conducted semi-structured interviews with 50 randomly selected students from Jilin University (comprising 30 males and 20 females) to ascertain their subgoals related to ordering only what they need and finishing the food they ordered. This study was conducted in accordance with all the principles of the Helsinki Declaration. Before the interview commenced, we informed the participants of the purpose of the interview: to understand the reasons for food waste in university canteens and the subgoals of ordering only what they need and finishing the food they ordered. We also advised them that the interview would last approximately 20–40 min, and that all their responses would be kept strictly confidential, with the survey results solely used for academic purposes. We assured them that the interview would not have any adverse effects on them. The interview only proceeded after the participants had signed the informed consent form.

The interviews revealed that the primary subgoals for ordering only what they need included keeping a healthy body, avoiding wasting money, protecting the environment, upholding social responsibility, reducing global hunger, and avoiding losing face from not finishing their food. The subgoals for finishing the food they ordered encompassed ensuring adequate intake of nutrients, avoiding wasting money, protecting the environment, upholding social responsibility, reducing global hunger, and avoiding losing face from not finishing their food. From the perspective of the goal system, these subgoals not only encompass the moral and social norms emphasized in prior research but also integrate individual goals, thereby providing more comprehensive content [13,15].

Subsequently, we assessed the strength of the subgoals–means cognitive association using a questionnaire survey. On 24 December 2020, we administered an online questionnaire to 199 students from Jilin University (see Supplementary Materials Section S1 for the “Order only what they need” questionnaire and Supplementary Materials Section S2 for the “Finish the food they ordered” questionnaire). Participants indicated their degree of agreement with the description of the cognitive association strength using a five-point Likert-type scale with 1 “strongly disagree” to 5 “strongly agree”; higher scores signified greater strength of cognitive association). We received 98 valid responses to the “Order only what they need” questionnaire, consisting of 49 male and 49 female respondents. For the questionnaire focused on finishing the food they ordered, we garnered 101 valid responses, of which 48 were from males and 53 were from females.

Research suggests that factors such as information processing and cognitive overload must be taken into account when crafting message content [59]. Excessive information can overwhelm individual attention and thereby diminish the efficacy of the intervention [60]. Accordingly, for both ordering and dining behavior, we chose three subgoals with the highest cognitive association, as well as a mutual focal goal of reducing food waste, to inform the design of relevant release-type messages (the strength of cognitive association between subgoals and means was ranked based on the mean (M) score obtained from the participants’ responses). The top three subgoals most cognitively associated with ordering only what they need were “keep a healthy body” (M = 3.66), “avoid wasting money” (M = 3.61), and “uphold social responsibility” (M = 3.57) (see Supplementary Materials Section S3, Table S1 for details). The top three subgoals most cognitively associated with finishing the food they ordered were “avoid wasting money” (M = 3.76), “ensure adequate intake of nutrients” (M = 3.56), and “uphold social responsibility” (M = 3.52) (see Supplementary Materials Section S3, Table S2 for details). Based on these findings, the goal system devised for reducing food waste in university canteens is illustrated in Figure 3.

In the current study, we delineate two distinct methodologies for approaching message frames. Specifically, gain-framed messages tend to be more effective when highlighting promotion goals, while loss-framed messages are generally more persuasive in emphasizing prevention goals [34,35]. According to our findings, reducing food waste and avoiding wasting money are prevention goals for which loss frames are more persuasive. Keeping a healthy body, ensuring adequate intake of nutrients, and upholding social responsibility are promotion goals that are more effectively addressed through gain frames. Drawing upon the release-type intervention messages’ design methodologies [13] and the corresponding message frames, we developed the following release-type intervention message at the ordering stage: “Order wisely to keep your body healthy and to uphold social responsibility. Failing to do so results in food waste and money loss. Please order only what you need” (see Supplementary Materials Section S4, Figure S1 for details). At the dining stage, the release-type intervention message was as follows: “Finish the food you ordered to ensure adequate intake of nutrients and to uphold social responsibility. Discarding food contributes to food waste and money loss. Please participate in the ‘Clean Your Plate’ Campaign” (see Supplementary Materials Section S4, Figure S2 for details).

3.4. The Feedback-Type Intervention Messages

This study employed weekly food waste data from the canteen at Jilin University to formulate feedback-type intervention messages [1]. Over a consecutive two-week period, researchers meticulously documented the number of food purchases and food waste conditions at the Nanling Fourth Canteen of Jilin University. Following the closure of the canteen each evening, both researchers and canteen staff collaboratively weighed the daily food waste. The cumulative food waste over the two-week period at the canteen amounted to 6880 kg, averaging about 3440 kg per week. The canteen mandated that all meal transactions be conducted through an electronic toll-collection system. Meal consumption data for this study were derived from the back-end records of the electronic toll-collection system. This study was conducted in accordance with all the principles of the Helsinki Declaration and data collection did not involve direct participation of canteen users. Throughout the two-week period, a total of 110,990 meals were consumed. Analysis of the data led to an estimation that each student wasted approximately 61.98 g of food per meal. This study utilized food waste data from Jilin University and another comparable university, referred to as University A, to design both upward and downward comparison feedback-type intervention messages [40]. Concurrently, the feedback-type messages in this study adopted a collective feedback approach. Incorporating waste quantity data, the feedback-type messages incorporated statements like “Let’s work together” to stimulate collective behavioral motivation [40]. The final versions of the feedback-type intervention messages were as follows:

The self-feedback-type intervention message read: “Based on weighing the waste every day, 3440 kg of food was wasted in this canteen last week, averaging 61.98 g per student per meal. Let’s work together!” (Specific data were presented in a visual bar graph. See Supplementary Materials Section S4, Figure S3).

The upward comparison feedback-type intervention message read: “Comparing to a comparable canteen in University A, this canteen generated more food waste last week. Let’s work together!” (Specific data were presented in a visual bar graph. Notes in small font at the bottom of the message image indicated that University A’s ranking and student population are similar to those of our university, and its comparable canteen’s size and variety of dishes are comparable to this one. See Supplementary Materials Section S4, Figure S4).

The downward comparison feedback-type intervention message read: “Comparing to a comparable canteen in University A, this canteen generated less food waste last week. Let’s work together!” (Specific data were presented in a visual bar graph. The characteristics of University A and the comparable canteen are as previously stated. See Supplementary Materials Section S4, Figure S5).

3.5. Fluency Test Results and the Pending Intervention Messages

On 11 March 2021, we conducted an online questionnaire survey to evaluate the semantic fluency of the five designed intervention messages, aiming to eliminate the potential influence of semantic fluency on the experimental outcomes [43]. This study was conducted in accordance with all the principles of the Declaration of Helsinki and all participants provided informed consent prior to participating in the study. The specific steps involved randomly assigning each participant to read an intervention message and subsequently answer five items related to semantic fluency. The survey received 474 responses, of which 459 were valid, comprising 187 males and 272 females. The distribution was as follows: 95 responses for “the release-type intervention message at the ordering stage”, 91 for “the release-type intervention message at the dining stage”, 85 for “self-feedback-type message”, 86 for “upward comparison feedback-type message”, and 102 for “downward comparison feedback-type message”.

As indicated by the data analysis in Supplementary Materials Section S5, the five test groups scored highly across the five dimensions of semantic fluency, and there were no significant disparities in the scores across these dimensions (all ps > 0.05). This indicates that the respondents had a clear understanding of the information, and the five messages exhibited consistent semantic fluency.

Because factors such as background color and position can differently capture participants’ attention and thus affect intervention outcomes, we controlled for these potential confounders during the message-design process. Mehta and Zhu [42] found that individuals prefer messages with a blue background over those with a red one. Thus, the intervention messages in this study were consistently designed with a blue background. Bialkova and van Trijp [41] noted that individuals respond more quickly when messages were placed in the top right corner. Therefore, we positioned the intervention messages in the top right corner of the ordering window during participants’ meal selection, and the messages at the dining stage were positioned at the top right corner of the dining table. The final messages are shown in Figures S1–S5.

3.6. Effectiveness Test of Pending Intervention Messages

3.6.1. Experiment Design

Control group: In recent years, China has advocated for “Practice Thrift and Opposition to Waste”, and canteens at Jilin University have already displayed related intervention messages (see Supplementary Materials Section S4, Figures S6 and S7). The content of these messages is primarily designed in a non-audience-oriented manner, mainly based on social norms, offering an opportunity for analyzing the disparities in intervention effects between non-audience-oriented and audience-oriented intervention messages. Therefore, the control group was designed in accordance with realistic situations to ensure that the effect of the information intervention was meaningful.

Portion size: Previous research indicated that appropriate food portions can reduce food waste [17,61]. Our study not only corroborated these findings but also revealed that participants have a preference for half portions (see Supplementary Materials Section S6; (in the pre-test of the effect of portion size, we adopted essentially the same experimental procedure as in the formal experiment to clarify the role of portion size, and at the same time verify the feasibility of the experimental procedure)). To control for potential confounding effects of portion size on the intervention outcomes, and to cater to participants’ preference for half portions, we controlled the portion size for the specific dishes in the pending intervention messages’ effectiveness experiment. Participants were given the option to choose either a half portion or a full portion of the dish, and they could freely decide the number of such portions to order.

Procedure: The presentation order of intervention messages influences the intervention effect [9]. Considering the ordering and dining scenarios among university students, the release-type intervention messages at the ordering stage and dining stage correspond to the actual consumption stage, so the presentation order in the experiment was also consistent. Since the feedback-type intervention messages reflect the final performance outcomes of both ordering and dining behaviors, and pertain to both, in our study we assessed the effects of presenting the feedback-type intervention messages at three different times: before the ordering stage, at the ordering stage, and at the dining stage. All participants were randomly assigned to 12 groups (see

Table 1. Summary table of experimental procedure.

Experiment Group Names (Group Number)	Experimental Procedure	The Intervention Messages in the Experimental Procedure	Number of Questionnaires
Control group (Group 1)	Step1-Step2-Step4-Step5-Step6-Step7	Step4: D6, Step6: D7	96
The release-type intervention message at the ordering stage (Group 2)	Step1-Step2-Step4-Step5-Step7	Step4: D1	92
Self-feedback-type intervention message at the ordering stage (Group 3)		Step4: D3	91
Upward comparison feedback-type intervention message at the ordering stage (Group 4)		Step4: D4	94
Downward comparison feedback-type intervention message at the ordering stage (Group 5)		Step4: D5	93
The release-type intervention message at the dining stage (Group 6)	Step1-Step2-Step5-Step6-Step7	Step6: D2	94
Self-feedback-type intervention message at the dining stage (Group 7)		Step6: D3	101
Upward comparison feedback-type intervention message at the dining stage (Group 8)		Step6: D4	104
Downward comparison feedback-type intervention message at the dining stage (Group 9)		Step6: D5	106
Self-feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 10)	Step1-Step2-Step3-Step4-Step5-Step6-Step7	Step3: D3, Step4: D1, Step6: D2	89
Upward comparison feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 11)		Step3: D4, Step4: D1, Step6: D2	90
Downward comparison feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 12)		Step3: D5, Step4: D1, Step6: D2	91

Notes: “Experimental procedure” indicates the corresponding experimental steps for each group. “The intervention messages in the experimental procedure” indicates which specific message was used by each group during which experimental step, and the content of these intervention messages can be found in Supplementary Materials Section S4. “Number of Questionnaires” indicates the quantity of questionnaires that were administered to each group.

), with Group 1 serving as the control group. The names of dishes, example images (taken from real scenarios), and prices used in the experimental scenarios were all designed based on actual conditions. This included 10 meat dishes, 10 vegetarian dishes, and 10 staple foods (see Supplementary Materials Section S7 for details). The survey used online questionnaires, and the experiment consisted of the following seven steps, but the intervention steps differed slightly among different groups (for specifics, see Table 1 and Figure 4):

Step 1: Participants were randomly assigned to either the intervention groups (see Groups 2–12 in Table 1) or the control group (see Group 1 in Table 1);

Step 2: Participants viewed a simulated meal-ordering scenario: “At midday on a day in November, after an 11:40 class, you depart from the academic building toward your usual canteen for lunch”;

Step 3: Participants were reminded: “As you walk past the canteen’s bulletin board, a message in the upper right corner catches your eye”. After reading this, participants proceeded to the next interface;

Step 4: Participants were reminded: “At the canteen’s ordering window, an intervention message was displayed in the top right corner of the glass showcase”. After reading this, participants proceeded to the next interface;

Step 5: Participants entered the ordering menu interface and placed their orders as needed. Food items in the online questionnaire were presented randomly to avoid sequence effects. After completing their orders, participants saw a display showing all selected foods, their images, their portion sizes, and prices. Upon confirmation, they proceeded to the next interface;

Step 6: Participants proceeded to dine at a table where another intervention message was posted in the top right corner. After reading the message, they moved on to the next interface;

Step 7: The questionnaire once again displayed all the selected food items, their images, portion sizes, and prices, followed by a question assessing the likelihood of participants finishing the food they ordered (“How likely are you to finish the food you ordered?” (0 = “waste it all”, 10 = “eat it all”)). Finally, participants completed sociodemographic questions, including gender, the degree you are pursuing, family location, and average monthly living expenses.

3.6.2. Participants

From 15 March to 15 April 2021, we conducted a face-to-face questionnaire survey of Jilin University students. This study was conducted in accordance with all the principles of the Declaration of Helsinki. All participants provided informed consent prior to participating in the study, and they were able to withdraw from the survey at any time without giving a reason. The researchers conducted sampling by selecting one individual randomly from every group of five encountered in densely populated areas such as the canteens, libraries, and playgrounds. Participants spent an average of 17.3 min on the experiment and they received a CNY 5 compensation after completing the experiment. We excluded those who took less than 9.8 min or spent more than CNY 27 (virtual expenditure of ordering meals in questionnaire), as these were more than three standard deviations from the mean [45]. After attention checks, 1141 valid questionnaires remained (see Table 1 for the number of valid questionnaires in each group). Among the participants, 562 were male, 826 were undergraduates, 821 resided in cities, 494 reported average monthly living expenses of less than CNY 1500, and 371 reported average monthly living expenses between CNY 1500 and 2000.

3.6.3. Results

The results showed that of 1,141 participants, only 198 did not choose the half-portion meal, indicating a preference for the half-portion option. Controlling for sociodemographic variables, multiple regression models revealed those not choosing the half portion selected fewer food items (M_{Non-Half Portion Selection} = 2.15) than those who did (M_{Half Portion Selection} = 3.44, β = 0.374, t = 13.508, p = 0.000).

Effect of information intervention on the ordering expenditure: In order to explore the effective intervention messages to reduce the ordering expenditure, we compared the experimental data from Groups 2, 3, 4, 5, 10, 11, and 12 (groups that provided intervention messages before ordering) with the data from the control group (Group 1). The independent variable we chose was information intervention, which was an eight-category variable. The control variables included gender (0 = Male; 1 = Female), the degree you are pursuing (Undergraduate = 0; Postgraduate = 1), family location (City = 0; Rural = 1), and average monthly living expenses (Average monthly living expenses < CNY 1500 = 1; CNY 1500 ≤ Average monthly living expenses ≤ CNY 2000 = 2; Average monthly living expenses > CNY 2000 = 3). The dependent variable, ordering expenditure, was normally distributed (i.e., skewness index < |3.00| and kurtosis index < |8.00|) [62,63]. The normality and homogeneity of residuals were tested and validated [64,65,66]. The Durbin–Watson value was 1.821, indicating that the residuals were considered to be independent [67]. All variance inflation factors (VIFs) tested for multicollinearity were found to be below the acceptable critical level of 10.0, ranging from 1.017 to 1.740, confirming that the data successfully passed the multicollinearity test [68]. We used multiple linear regression to test whether there was a significant difference between the experimental groups and the control group in information intervention effect. The regression model is as follows:

$$Y={\beta }_0+{\sum }_{j=1}^7{\beta }_j {information}_j+{\beta }_8 \mathit{gender}+{\beta }_9 \mathit{degree}+{\beta }_{10} \mathit{location}+{\beta }_{11} {\mathit{expense}}_1+{\beta }_{12} {\mathit{expense}}_2+\epsilon$$

(1)

where “Y” represents “ordering expenditure”, which is a continuous variable, “information_j” is a dummy variable formed from “intervention information” with the control group serving as the reference group, “gender” represents the gender of the participants, “degree” represents “the degree you are pursuing”, “location” represents “family location”, “average monthly living expenses” takes “less than CNY 1500” as the reference group to form two dummy variables expense₁ and expense₂, β₀ denotes the intercept term, β₁ through β₁₂ are the regression coefficients to be estimated in this study, utilizing the Ordinary Least Squares (OLS) estimation method, and ε represents the error term.

Controlling for sociodemographic variables, the results of the multiple linear regression (see

Table 2. Effect sizes per intervention message.

		Model 1 (Ordering Expenditure)						Model 2 (Expected Amount of Food Waste)
		β	t	p	β	t	p	β	t	p	β	t	p
Ordering stage	The release-type intervention message at the ordering stage (Group 2)	−0.092	−1.921	0.055	−0.101	−2.149	0.032	−0.107	−2.714	0.007	−0.104	−2.649	0.008
	Self-feedback-type intervention message at the ordering stage (Group 3)	−0.061	−1.273	0.203	−0.066	−1.398	0.163	−0.109	−2.755	0.006	−0.104	−2.642	0.008
	Upward comparison feedback-type intervention message at the ordering stage (Group 4)	−0.087	−1.813	0.070	−0.089	−1.890	0.059	−0.103	−2.598	0.010	−0.100	−2.518	0.012
	Downward comparison feedback-type intervention message at the ordering stage (Group 5)	−0.118	−2.451	0.014	−0.134	−2.847	0.005	−0.123	−3.097	0.002	−0.117	−2.960	0.003
Dining stage	The release-type intervention message at the dining stage (Group 6)							−0.038	−0.955	0.340	−0.032	−0.799	0.425
	Self-feedback-type intervention message at the dining stage (Group 7)							−0.047	−1.171	0.242	−0.047	−1.184	0.237
	Upward comparison feedback-type intervention message at the dining stage (Group 8)							−0.004	−0.090	0.929	−0.005	−0.124	0.901
	Downward comparison feedback-type intervention message at the dining stage (Group 9)							−0.077	−1.883	0.060	−0.077	−1.914	0.056
Before ordering stage + Ordering stage + Dining stage (Mixed intervention)	Self-feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 10)	−0.127	−2.664	0.008	−0.123	−2.630	0.009	−0.102	−2.601	0.009	−0.103	−2.646	0.008
	Upward comparison feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 11)	−0.127	−2.662	0.008	−0.126	−2.690	0.007	−0.097	−2.459	0.014	−0.097	−2.468	0.014
	Downward comparison feedback-type intervention message before the ordering stage + the release-type intervention message at the ordering stage + the release-type intervention message at the dining stage (Group 12)	−0.154	−3.217	0.001	−0.140	−2.986	0.003	−0.096	−2.450	0.014	−0.104	−2.648	0.008
	Gender 1				−0.203	−5.619	0.000				0.115	3.907	0.000
	The degree you are pursuing 2				0.008	0.232	0.817				0.006	0.219	0.827
	Family location 3				0.037	0.985	0.325				−0.010	−0.312	0.755
	CNY 1500 ≤ Average monthly living expenses ≤ CNY 2000 4				0.056	1.405	0.161				0.017	0.511	0.609
	Average monthly living expenses > CNY 2000 4				0.143	3.552	0.000				0.019	0.568	0.570
	F	2.050 *			4.973 ***			2.395 **			2.671 ***
	Adj-R2	0.010			0.061			0.013			0.023

Notes: n_model-1 = 736. n_model-2 = 1141. Reporting standardized regression coefficients. * p < 0.05. ** p < 0.01. *** p < 0.001. Reference group is control group (Group 1). ¹ (0 = Male; 1 = Female). ² (Undergraduate = 0; Postgraduate = 1). ³ (City = 0; Rural = 1). ⁴ (Average monthly living expenses < CNY 1500 = 1; CNY 1500 ≤ Average monthly living expenses ≤ CNY 2000 = 2; Average monthly living expenses > CNY 2000 = 3). In addition, significant groups are highlighted in bold, and marginal significance groups are in italics.

and Figure 5) indicated that the intervention effects of five groups exceeded the control group. The mean ordering expenditure of the control group was CNY 13.30. The intervention effects of the three mixed intervention groups, including Group 10 (β = −0.123, t = −2.630, p = 0.009), Group 11 (β = −0.126, t = −2.690, p = 0.007), and Group 12 (β = −0.140, t = −2.986, p = 0.003) were all better than those of the control group. The intervention effects of both the release-type intervention message (Group 2 (β = −0.101, t = −2.149, p = 0.032)) and downward comparison feedback-type intervention message at the ordering stage (Group 5 (β = −0.134, t = −2.847, p = 0.005)) were significantly superior to those of the control group. However, the intervention effects of both the self-feedback-type intervention message (Group 3 (β = −0.066, t = −1.398, p = 0.163)) and upward comparison feedback-type intervention message at the ordering stage (Group 4 (β = −0.089, t = −1.890, p = 0.059)) did not differ from those of the control group. The intervention effects of the five experimental groups that significantly outperformed the control group, ranked from highest to lowest, were as follows: Group 12 (β = −0.140, t = −2.986, p = 0.003), Group 5 (β = −0.134, t = −2.847, p = 0.005), Group 11 (β = −0.126, t = −2.690, p = 0.007), Group 10 (β = −0.123, t = −2.630, p = 0.009), and Group 2 (β = −0.101, t = −2.149, p = 0.032). Notably, among these, Group 12 exhibited the most significant effect on ordering expenditure. Compared with the results before adding control variables, the intervention effect of the release-type intervention message at the ordering stage was slightly improved, and the significance of other intervention messages did not change. We confirmed that the results of this study were stable.

Effect of information interventions on the expected amount of food waste: Due to the difficulty in directly summing the quantities of different dishes for inter-group comparisons, this study introduced a comprehensive index—the expected amount of food waste—to measure total food waste.

The expected amount of food waste = (1 − likelihood of finishing score × 0.1) × ordering expenditure. A higher score for the expected amount of food waste indicates a greater degree of waste.

In this study, the independent variable was information intervention, which is a 12-category variable. The control variables include gender (0 = Male; 1 = Female), the degree you are pursuing (Undergraduate = 0; Postgraduate = 1), family location (City = 0; Rural = 1), and average monthly living expenses (Average monthly living expenses < CNY 1500 = 1; CNY 1500 ≤ Average monthly living expenses ≤ CNY 2000 = 2; Average monthly living expenses > CNY 2000 = 3). The dependent variable, expected amount of food waste, was normally distributed (i.e., skewness index <|3.00|and kurtosis index < |8.00|) [62,63]. The normality and homogeneity of residuals were tested and validated [64,65,66]. The Durbin–Watson value was 2.020, indicating that the residual was considered to be independent [67]. The variance inflation factor (VIF) method was used to judge multicollinearity. The results showed that the multicollinearity test was passed because all VIFs were less than 10 (VIF values between 1.015 and 1.913) [68]. We used multiple linear regression to test whether the expected amount of food waste in the experimental groups was significantly different from that in the control group. The regression model is as follows:

$$Y={\beta }_0+{\sum }_{j=1}^{11}{\beta }_j {information}_j+{\beta }_{12} \mathit{gender}+{\beta }_{13} \mathit{degree}+{\beta }_{14} \mathit{location}+{\beta }_{15} {\mathit{expense}}_1+{\beta }_{16} {\mathit{expense}}_2+\epsilon$$

(2)

where “Y” represents “expected amount of food waste”, which is a continuous variable, “information_j” is a dummy variable formed from “intervention information” with the control group serving as the reference group, “gender” represents the gender of the participants, “degree” represents “the degree you are pursuing”, “location” represents “family location”, “average monthly living expenses” takes “less than CNY 1500” as the reference group to form two dummy variables expense₁ and expense₂, β₀ denotes the intercept term, β₁ through β₁₆ are the regression coefficients to be estimated in this study, utilizing the Ordinary Least Squares (OLS) estimation method, and ε represents the error term.

Controlling for sociodemographic variables, the multiple linear regression results (refer to Table 2 and Figure 6) showed that seven intervention groups significantly outperformed the control group. The mean expected amount of food waste for the control group was CNY 2.17. The intervention effects of the release-type intervention message (Group 2 (β = −0.104, t = −2.649, p = 0.008)), self-feedback-type intervention message (Group 3 (β = −0.104, t = −2.642, p = 0.008)), upward comparison feedback-type intervention message (Group 4 (β = −0.100, t = −2.518, p = 0.012)), and downward comparison feedback-type intervention message at the ordering stage (Group 5 (β = −0.117, t = −2.960, p = 0.003)) were significantly superior to those of the control group. The intervention effects of the three mixed intervention groups, including Group 10 (β = −0.103, t = −2.646, p = 0.008), Group 11 (β = −0.097, t = −2.468, p = 0.014), and Group 12 (β = −0.104, t = −2.648, p = 0.008) were all better than those of the control group. On the contrary, the intervention effects of the release-type intervention message (Group 6 (β = −0.032, t = −0.799, p = 0.425)), self-feedback-type intervention message (Group 7 (β = −0.047, t = −1.184, p = 0.237)), upward comparison feedback-type intervention message (Group 8 (β = −0.005, t = −0.124, p = 0.901)), and downward comparison feedback-type intervention message at the dining stage (Group 9 (β = −0.077, t = −1.914, p = 0.056)) did not differ from those of the control group.

The intervention effects of the seven experimental groups that significantly outperformed the control group, ranked from highest to lowest, were as follows: Group 5 (β = −0.117, t = −2.960, p = 0.003), Group 2 (β = −0.104, t = −2.649, p = 0.008), Group 3 (β = −0.104, t = −2.642, p = 0.008), Group 12 (β = −0.104, t = −2.648, p = 0.008), Group 10 (β = −0.103, t = −2.646, p = 0.008), Group 4 (β = −0.100, t = −2.518, p = 0.012), and Group 11 (β = −0.097, t = −2.468, p = 0.014). Among them, Group 5 had the best intervention effect. Compared with the results before adding control variables, the significance of information intervention did not change, and we confirmed that the results of this study were stable.

4. Overall Discussion and Implication

4.1. Overall Discussion

We implemented an audience-oriented approach to reduce food waste in Chinese university canteens, taking into account factors like portion size and presentation order. Moreover, our findings are in line with earlier research, revealing that smaller portion sizes not only mitigate food waste [17,61] but also enrich the meal choices available to students [15].

To reduce food waste in university canteens, it is recommended to prioritize the downward comparison feedback-type messages at the ordering stage. Such messages have been found to not only reduce the expected amount of food waste and expenditure but also enrich the variety of food items selected. Research by Dupré and Meineri [40] suggested that comparative feedback-type messages significantly influence waste-recycling behavior in university canteens, with downward comparisons generally more effective than upward ones. Our study’s results affirm that downward comparison feedback outperforms upward comparison feedback in terms of economic and aggregate metrics. This likely occurs because downward comparisons improve self-evaluation and foster positive feelings [37].

The release-type intervention message at the ordering stage notably influenced both the ordering expenditure and the expected amount of food waste. Only the downward comparison feedback-type message at the ordering stage surpassed its effectiveness in reducing the expected amount of food waste. The effect of the self-feedback-type message at the ordering stage on reducing the expected amount of food waste was similar to the release-type intervention message at that stage. This was consistent with studies conducted by Whitehair, Shanklin, and Brannon [13] and Pinto, dos Santos Pinto, Silva Melo, Campos, and Cordovil [14]. Whitehair, Shanklin, and Brannon [13] found that both the release-type intervention message and the self-feedback-type message markedly reduced university canteen food waste, and their effects were almost on par with each other. Pinto, dos Santos Pinto, Silva Melo, Campos, and Cordovil [14] suggested that the release-type intervention message lowered the waste index. However, our study conflicted with the findings of Visschers, Gundlach, and Beretta [17], who found the release-type intervention message ineffective. This may be because the message was not tailored to the target audience, which might reduce the feasibility of the information, thereby reducing the perceived behavioral control (PBC) [17].

The three mixed messages in this study had favorable intervention effects, and a downward comparison feedback-type intervention message at the ordering stage can achieve the same impact. When conditions are similar, we recommend a downward comparison feedback-type intervention message at the ordering stage as it streamlines the intervention process and is more cost-effective.

Regarding presentation order, when it comes to food waste it is best to intervene before and at the ordering stage. The effectiveness of the intervention at the dining stage was not satisfactory in our study. Qian, Li, Cao, Wang, and Jin [15] also encountered the same issue. They noted that the widely-used “Clean Your Plate” messages on dining tables in China were not effective in reducing food waste. A possible reason is that participants experienced pronounced hunger before the meal, which was subsequently alleviated by eating [9]. Hungry individuals focus their visual attention more on food-related stimuli [69], and visual attention played a role in decision-making processes [70]. Therefore, the differences in hunger levels between the ordering and dining stages, coupled with the interconnectedness of hunger, visual attention, and decision-making [70], may contribute to the unfavorable intervention effects at the dining stage.

Overall, our study confirmed that, except at the dining stage, audience-oriented intervention messages significantly outperformed traditional messages (i.e., the control group, which used intervention messages based on social norms). This substantiates the view that an audience-oriented approach is a viable strategy for promoting positive behavior change.

4.2. Implication

4.2.1. Theoretical Implications

First, communication is regarded as a continuous cycle with message meaning emerging from the interplay among messages, participants (senders and receivers), and contextual factors [20]. Information intervention is essentially a process of communication, and crafting effective intervention messages requires a profound understanding of the cognitive, motivational, and historical behavioral performance of the target audience. Previous studies on food waste informational interventions have largely centered on presentation method (like communication modality), while often overlooking the tailored nature of the content. According to the elaboration likelihood model (ELM), when the audiences perceive the information as personally relevant, the information undergoes deeper and more meticulous processing [18].

This study was guided by the goal systems theory and dynamically interpreted individual behavior from the perspective of “motivation as cognition”. Concurrently, by combining the feedback paradigms and integrating the past behavior of the target audience, an audience-oriented food waste behavior information intervention approach was established. This approach enhances the correlation between messages and the target audience, and enriches the theoretical achievements in the field of information intervention. This paper was the first attempt to apply the goal systems theory to the field of consumer communication, expanding the application scenarios of this theory and providing new theoretical insights for the design of communication messages. In addition, the audience-oriented information intervention approach can be applied to other extensive audience behavior interventions, especially in the reverse marketing field where goal system cognition is relatively simple and consistent, such as green consumption and low-carbon behavior promotion.

Our study is a specific application of the goal systems theory, in which we emphasize that individuals’ choice of means can be understood through the relationship between individual goals and means [28]. We activated the subgoals corresponding to a certain means through information intervention, expanded the size of the multifinality set, improved the perceived value of the means, and urged consumers to choose that particular means [28,29]. Our findings indicate that the application of the goal systems theory in designing release-type intervention messages has indeed achieved remarkable intervention effects. This not only validates the underlying principles of the goal systems theory and expands its application scenarios, but also provides new theoretical insights for the design of intervention messages.

Second, previous research has found that information intervention yields divergent outcomes in reducing food waste, so it is necessary to consider more details, such as message type, when exploring the effectiveness of information intervention [17]. To address the limitations observed in prior research, where a comprehensive examination of intervention message types was lacking, there was limited utilization of message types, and there was only a weak correlation with the presentation order, this study comprehensively considered the release-type intervention messages and feedback-type intervention messages (including a self-feedback-type intervention message, an upward comparison feedback-type intervention message, and a downward comparison feedback-type intervention message), and grouped the information intervention according to the presentation order when different messages appeared. By comparing with the control group, this study identified the most efficacious intervention message, thus making a noteworthy contribution to the realm of social comparison within the context of food waste.

4.2.2. Managerial Implications

This study offers meaningful management implications to the managers of canteens and restaurants.

First, our results confirm the important influence of portion size on reducing food waste. Canteens and restaurants should provide customers with a diverse range of portion sizes to choose from, which not only encourages customers to participate in initiatives aimed at curbing food waste but also aligns with the overarching objective of sustainable development [7], enhances the nutritional content of the food items, and encourages healthier eating practices.

Second, in order to reduce food waste, restaurants and canteens typically post intervention messages at multiple locations, such as placing posters throughout the restaurants and canteens, and posting intervention messages at the ordering counter and on dining tables. The content of these messages is also diverse, with some emphasizing environmental hazards, while others provide simple reminders to customers to reduce food waste. In light of the findings from this study, it is evident that while the aforementioned intervention method exhibits efficacy in mitigating food waste, a more economically efficient alternative is discernible. For instance, our study showed that the downward comparison feedback-type message at the ordering stage was more effective than other messages. Under the premise of limited budget, priority can be given to showing the downward comparison feedback-type intervention message to customers at the ordering stage, or the message can be conveyed by the waiter verbally, so that customers can actively engage in the food waste reduction campaign. Furthermore, our results indicated that intervention messages at the dining stage lacked effectiveness. Restaurants should avoid focusing on intervening only at the dining stage. Instead, they should emphasize encouraging customers to order wisely, thereby reducing food waste.

4.3. Limitations and Future Research

In constructing the contextual goal systems, we used a survey-based approach. It only took into account explicit goals (addressing both normative and personal goals), while background goals can also influence behavioral decisions unconsciously [28]. The neglect of background goals beyond consciousness is a significant limitation of the theoretical approach proposed in this study.

This study surveyed only university students dining in canteens. The applicability of the audience-oriented intervention approach to other samples needs further validation. Furthermore, while our study confirmed the effectiveness of information interventions in reducing food waste, it did not delve into the underlying mechanisms. Subsequent research may consider using the cognitive dissonance theory to explain the mechanism of informational intervention [71]. When confronted with food waste, numerous individuals may inherently adopt the cognition of “Do not care about food waste” due to habitual influence. However, upon exposure to information intervention advocating against food waste, a shift in cognition may emerge: “More willingness to follow the food waste reduction campaign” [9]. The disparity between these two cognitive stances can induce cognitive dissonance, culminating in psychological discomfort [72]. To mitigate this discomfort, individuals might either opt to change their behavior, actively participating in food waste reduction initiatives, or they might change their cognition and rationalize their food waste behavior. This may elucidate why appropriate information intervention can effectively reduce food waste: it offers individuals a psychological incentive to alter their behavior.

5. Conclusions

Our study aimed to identify the most effective intervention messages to reduce food waste. To enhance the effectiveness of the messages, we developed an audience-oriented approach in formulating release-type and feedback-type intervention messages. Release-type messages were grounded in the goal systems theory, mirroring the audience’s cognition and motivations related to food waste behavior. Feedback-type messages predominantly highlighted the food waste quantity and trend of both target and comparative audiences over a specific period. In the study, we comprehensively considered message frames, comparison paradigm, attention-grabbing elements like font and color, and intervention timing. The approach ensured that the content of the messages was tailored to the target audience, enhancing the messages’ relevance. Tailored intervention messages are more attention-grabbing than non-tailored, compelling the audience to engage more deeply and enhancing the messages’ persuasive impact. We applied this approach by intervening in food waste behavior among Chinese university students in canteens. Our findings indicated that intervention messages tailored to the target audience had a better effect. Compared to the control group, the downward comparative feedback-type message at the ordering stage was the most effective in the university canteen scenario. It not only significantly reduced the ordering expenditure but also considerably lowered the expected amount of food waste. Under similar conditions, prioritizing this message is advisable. Our study underscores the imperative of audience-oriented message formulation in interventions. The approach we introduced contributes significantly to the existing research on information intervention. Furthermore, it broadens the application of the social comparison theory in the context of food waste.