📄 agriculture

Asymmetric Effects of Fishing Attributes on Angler Satisfaction in Chinese Fishing Gardens

This study analyzes online review data from Chinese fishing gardens using advanced machine learning and Kano modeling to reveal that angler satisfaction is asymmetrically driven by catch-related attributes—particularly fish size and quantity—which exhibit threshold effects and complex interactions with non-catch factors like water quality and cost.

Original authors: Guiying Chen, Zhengyong Yang, Zhe Chen, Qilei Zhao, Xinjun Chen

Published 2026-07-23

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Guiying Chen, Zhengyong Yang, Zhe Chen, Qilei Zhao, Xinjun Chen

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are the chef of a very special restaurant where the main dish isn't a steak or a pizza, but the thrill of catching a fish. This isn't just about eating; it's a whole social event where people pay to spend their day by the water, hoping to hook something. Scientists who study this are called recreational fisheries researchers. They want to know: what makes a fisherman happy? Is it just the size of the fish? Is it the price of the ticket? Or is it how nice the waiter is?

To answer this, researchers often use a concept called the "Kano model." Think of this like a video game with different types of power-ups. Some power-ups are "Must-Haves": if you don't have them, you lose immediately (like having no health bar), but having them doesn't make you super-powerful; it just keeps you in the game. Other power-ups are "Attractive": you don't expect them, but if you get one, you feel like a total hero. Then there are "One-Dimensional" power-ups: the more you get, the happier you are, but the less you get, the sadder you are. The big question this paper asks is: do these rules work differently when things go wrong compared to when things go right? Does a bad experience hurt more than a good experience feels good?

This study dives into the world of "Chinese fishing gardens." These are commercial ponds where people pay to fish, and the managers stock the water with fish to ensure there's something to catch. The researchers didn't just ask people to fill out boring surveys; they acted like digital detectives. They scraped over 49,000 real online reviews from Chinese fishing apps, using advanced computer programs to read the text and figure out what people were actually saying. They used a smart algorithm (XGBoost) to predict satisfaction and a special tool (SHAP) to see exactly which words in the reviews made people happy or angry. They then mapped these findings onto their "Kano model" to see which fishing attributes were the "Must-Haves" and which were the "Attractive" bonuses.

Here is what they found, and it's a bit like a rollercoaster of emotions. First, the most important thing for a fisherman in these gardens is the catch. Specifically, the size of the fish and the number of fish they catch. These two factors alone accounted for about 63.06% of the total impact on satisfaction. But here is the twist: they act very differently.

Fish Size is an "Attractive" attribute. Think of it like finding a golden ticket in your chocolate bar. If you catch a tiny fish, you aren't necessarily furious; you just don't get the "wow" factor. But if you catch a massive, trophy-sized fish, your satisfaction skyrockets. It's a bonus that makes the day legendary.

Fish Quantity, however, is a "Must-Have" attribute. This is the health bar. If you go fishing and catch nothing or very few fish, you are instantly devastated. The negative feeling of an empty net is the strongest negative force in the study. But if you catch a lot of fish, it doesn't make you super happy; it just makes you satisfied enough to not be angry. You need a decent amount of fish just to feel like the trip was worth it.

Then there is Fish Species. This is the "One-Dimensional" attribute. It's a straight line: the more you like the type of fish, the happier you are; the less you like it, the sadder you are. It's balanced, not extreme.

The study also looked at things other than the fish. Water Quality and Cost (the price you pay) turned out to be "Must-Haves" too. If the water looks dirty or the price feels unfair, your satisfaction crashes. But if the water is crystal clear or the price is a great deal, it doesn't boost your happiness as much as a bad experience hurts it. Interestingly, Manager Charisma (how friendly and helpful the staff is) is an "Attractive" attribute. A grumpy manager won't ruin your day, but a super-friendly, charismatic manager can turn a good day into a great one.

The researchers also discovered something called "thresholds." Imagine a light switch that is hard to flip. For positive things (like catching a big fish), you need to cross a high bar before you feel the extra joy. But for negative things (like bad water quality), the switch flips very easily; even a small drop in quality can make you unhappy.

Finally, they looked at how these factors mix together. When you have lots of good things happening (big fish AND many fish), they actually start to cancel each other out a little bit; you can't get happier than "very happy." But when you have lots of bad things (dirty water AND high cost), the misery doesn't double; it kind of levels off because you're already at the bottom. However, if you catch a ton of fish, it can actually fix a bad feeling about the price. It's like saying, "Well, the ticket was expensive, but I caught so much fish that it was worth it!"

In short, for these fishing gardens, the managers need to make sure the "Must-Haves" (enough fish, clean water, fair price) are never missing, because missing them causes the most pain. Then, they should try to create "Attractive" moments (huge fish, friendly staff) to give people that extra spark of joy. The study suggests that catching fish isn't just about biology; it's a complex mix of psychology, where the fear of an empty net is much stronger than the joy of a full one.

Technical Summary: Asymmetric Effects of Fishing Attributes on Angler Satisfaction in Chinese Fishing Gardens

Problem Statement

Recreational fisheries function as complex social-ecological systems where management effectiveness relies heavily on angler satisfaction. While existing literature identifies catch-related and non-catch-related attributes as determinants of satisfaction, current research often treats these relationships as linear or symmetric. There is a significant gap in understanding:

Asymmetry: Whether positive and negative manifestations of the same attribute exert unequal effects on satisfaction (e.g., does the absence of a feature cause more dissatisfaction than its presence causes satisfaction?).
Thresholds: Whether attribute effects are activated only after reaching specific performance levels.
Interactions: How multiple attributes interact to shape satisfaction evaluations.
Contextual Specificity: Empirical evidence remains limited for transitional economies like China, where commercial "fishing gardens" (fee-based, stocked ponds) represent a rapidly growing sector distinct from traditional wild fisheries or Western "put-and-take" models.

Methodology

The study employs a mixed-method computational approach integrating text mining, machine learning, and behavioral modeling to analyze large-scale, unstructured online review data.

1. Data Collection and Preprocessing

Source: 49,837 valid reviews were scraped from two major Chinese angling platforms ("Angler" app and "Fishing Community" web platform) on June 1, 2024.
Filtering: Unstructured noise, private information, and commercial promotions were removed from an initial pool of ~101,000 comments.

2. Attribute Identification and Sentiment Analysis

Topic Modeling: Latent Dirichlet Allocation (LDA) was used to identify 10 distinct angling attributes (e.g., fish size, quantity, species, water quality, cost, manager charisma).
Syntactic Parsing: The Baidu DDParser dependency parser segmented text into 227,754 clauses for fine-grained matching.
Sentiment Scoring: SnowNLP (a Chinese NLP library) calculated sentiment probabilities for both complete reviews (serving as the dependent variable for satisfaction) and individual clauses (extracting positive and negative sentiment features for specific attributes).

3. Predictive Modeling: XGBoost–SHAP

Algorithm: An XGBoost (Extreme Gradient Boosting) model was trained to predict review sentiment probability based on attribute-specific sentiment features.
Optimization: Hyperparameters were tuned using Bayesian optimization (TPE algorithm) to maximize $R^2$ and minimize RMSE.
Interpretability: SHAP (SHapley Additive exPlanations) values were computed to decompose predictions into additive feature contributions. Crucially, each attribute was decomposed into separate positive ("+") and negative ("−") sentiment features to capture asymmetric effects without signal cancellation.

4. Classification Framework: Improved Kano Model

Adaptation: An improved Kano model was developed to translate SHAP values into managerial attribute categories (Attractive, Must-have, One-dimensional, Indifferent).
Metrics:
- Impact Index ( $\gamma_j$ ): Measures overall impact intensity based on the Euclidean distance of positive and negative SHAP values.
- Asymmetry Index ( $\lambda_j$ ): The ratio of positive to negative SHAP impact.
Classification Rules: Attributes were categorized based on thresholds for $\gamma$ (impact intensity) and $\lambda$ (asymmetry), allowing for the identification of "must-have" (strong negative impact, weak positive) vs. "attractive" (strong positive impact, weak negative) attributes directly from unstructured text.

Key Results

1. Attribute Importance and Asymmetry

Dominance of Catch: Catch-related factors accounted for 63.06% of the total impact index.
Fish Size (Attractive): Classified as an attractive attribute ( $\lambda > 1.6$ ). It is the strongest positive feature; its presence significantly boosts satisfaction, but its absence does not cause equivalent dissatisfaction.
Fish Quantity (Must-have): Classified as a must-have attribute ( $\lambda < 0.8$ ). It is the strongest negative feature; insufficient quantity triggers pronounced dissatisfaction, while sufficient quantity does not continuously elevate satisfaction.
Fish Species (One-dimensional): Exhibits relatively balanced positive and negative effects ( $0.8 \le \lambda \le 1.6$ ).
Non-Catch Factors:
- Water Quality & Cost: Classified as must-have attributes. Negative sentiment regarding these factors strongly reduces satisfaction.
- Manager Charisma: Classified as an attractive attribute. Positive interpersonal interactions significantly enhance satisfaction.

2. Threshold Characteristics

Positive Features: Require high activation levels to generate satisfaction gains. For example, positive features for fish size, quantity, and species only enhance satisfaction after reaching specific thresholds (e.g., fish size+ threshold at 0.092).
Negative Features: Trigger dissatisfaction at lower levels. Deteriorating water quality, for instance, can trigger dissatisfaction at very low thresholds (0.012).

3. Interaction Patterns

Interaction Strength: Strongest interactions occur within catch-related attributes and between catch attributes and water quality/cost.
High-High Interactions:
- Positive Features: High-level positive features (e.g., large fish + many fish) tend to show negative interactions (marginal substitution). Once one positive attribute exceeds the reference point, the marginal gain from another positive attribute diminishes.
- Negative Features: High-level negative features (e.g., poor water quality + high cost) tend to show positive interactions (attenuation of marginal dissatisfaction). When an experience is already poor, additional negative factors add less marginal dissatisfaction.
Cost Moderation: Catch returns (quantity/size) moderate cost dissatisfaction. Abundant or large catches can offset negative perceptions of cost.

Significance and Claims

The paper claims to provide empirical evidence for satisfaction-oriented management in recreational fisheries by moving beyond linear assumptions.

Methodological Contribution: The study demonstrates the viability of integrating LDA, sentiment analysis, XGBoost–SHAP, and an improved Kano model to analyze unstructured online review data, bypassing the limitations of traditional survey-based Kano studies.
Theoretical Insight: It validates Prospect Theory and the Kano model in a recreational fishing context, confirming that anglers are more sensitive to losses (e.g., low catch, poor water) than gains, and that attribute effects are non-linear and threshold-dependent.
Management Implications:
- Priority Setting: Managers should prioritize preventing "must-have" failures (insufficient fish quantity, poor water quality, high cost) as these are the primary drivers of dissatisfaction.
- Differentiated Strategies: "Attractive" attributes (fish size, manager charisma) should be leveraged to create memorable, positive experiences but are secondary to ensuring baseline "must-have" performance.
- Contextual Nuance: In Chinese fishing gardens, relational service (manager charisma) plays a critical role, reflecting local cultural dynamics (guanxi), while species composition acts more as a pre-trip choice cue than a post-trip satisfaction driver.

The authors modestly note limitations regarding self-selection bias in online reviews and the lack of stratified analysis across China's diverse regions, suggesting future work should integrate multiple data sources and compare commercial gardens with natural waters and Western systems.