📄 social_science

A Network Analysis of Loneliness in Intimate Relationships and AI Attachment Among Married Saudi Couples

This study utilizes network analysis on 306 married Saudi adults to reveal that emotional hurt is the central driver of marital loneliness, which weakly connects to AI attachment through social substitution, suggesting that therapeutic efforts should prioritize repairing emotional wounds within the marriage rather than directly targeting AI use.

Original authors: Fatma Khalifa Elsayed, Jwanaa Abdullah Aljohani, Khawla farid Obaidullah, Dalia Khalid Kaki, Randa Adham Jumah, Roa Talal Sagga

Published 2026-07-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Fatma Khalifa Elsayed, Jwanaa Abdullah Aljohani, Khawla farid Obaidullah, Dalia Khalid Kaki, Randa Adham Jumah, Roa Talal Sagga

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

The Big Picture: A Digital Safety Net vs. A Leaky Boat

Imagine a married couple on a boat. Sometimes, the boat develops a leak, and the passengers feel cold, isolated, and hurt, even though they are sitting right next to each other. This feeling is called marital loneliness.

In the past, people might have tried to fix the leak or talk to a friend. But today, there is a new option: Artificial Intelligence (AI). Think of AI as a very polite, always-available robot companion that never gets angry, never leaves, and always says what you want to hear.

This study asked: When people feel lonely in their marriage, do they turn to this robot companion? And does the robot help fix the leak, or does it make the boat feel even more unstable?

Who Was Studied?

The researchers looked at 306 married people in Saudi Arabia.

Most were women (about 3 out of 4).
Most were over 35 years old.
Most had been married for a long time (over 8 years).
They all answered questions about how close they feel to their spouse, how lonely they feel, and how attached they are to AI tools.

The Two Main Tools Used

To understand these feelings, the researchers used two "measuring tapes" (scales) that they tested and confirmed work well in Arabic:

The AI Attachment Tape: This measures how much someone bonds with AI. It looks at three things:
- Emotional Closeness: Do you feel like the AI is a friend?
- Social Substitution: Do you use the AI instead of talking to people?
- Moral Appreciation: Do you think the AI has a "soul" or moral value?
The Marital Loneliness Tape: This measures the specific pain of being lonely while married. It looks at three things:
- Detachment: Feeling emotionally distant from your spouse.
- Emotional Hurt: Feeling deeply wounded by the relationship.
- Guilt: Feeling bad about the state of your marriage.

The Main Discovery: The "Hub" of the Problem

The researchers used a special computer method (called Network Analysis) to map out how all these feelings connect to each other. Imagine a spiderweb where every feeling is a knot, and the strings connecting them show how strongly they influence each other.

The most important finding:
The knot labeled "Emotional Hurt" was the strongest, most central part of the entire web.

The Metaphor: Think of "Emotional Hurt" as the heavy anchor in the middle of a ship. If the anchor is heavy, it pulls everything else down with it.
What it means: When a person feels hurt in their marriage, it almost instantly triggers feelings of guilt and detachment. It also seems to push them toward using AI as a substitute for human connection. The study found that "Emotional Hurt" is the engine driving the whole cycle.

The Surprising Twist: The Robot Makes the Pain Worse

The study found a strange and sad pattern regarding AI Social Substitution (using AI to replace human interaction).

The Metaphor: Imagine you have a stomach ache. Instead of taking medicine, you eat a candy bar that tastes good for a second but actually makes your stomach ache worse an hour later.
What the paper says: People who used AI to replace their spouse didn't feel better. Instead, using the AI was strongly linked to more guilt and more emotional hurt. The AI didn't fix the loneliness; it seemed to deepen the pain of the marriage.

What About the Other Feelings?

AI Moral Appreciation: This was the "loneliest" knot in the web. Thinking that AI has moral value or a soul didn't really connect to the marital pain. It was a side note, not a main driver.
The Two Groups: The study confirmed that "AI Attachment" and "Marital Loneliness" are two separate groups of feelings (like two different rooms in a house). However, there is a small door between them. The strongest door connects using AI as a substitute to feeling guilty and hurt in the marriage.

The Conclusion

The paper concludes that the relationship between marital loneliness and AI isn't a simple "I'm lonely, so I talk to a robot" story. It's a cycle:

You feel hurt in your marriage.
This hurt makes you feel guilty and distant.
You turn to AI to fill the gap.
Using AI as a substitute actually makes the hurt and guilt feel even stronger, locking you in a loop.

The Bottom Line:
The study suggests that if you want to stop this cycle, you shouldn't just focus on banning the AI. You have to fix the Emotional Hurt first. If you heal the hurt in the marriage, the need to use the robot as a crutch, and the feelings of guilt, will likely go away on their own.

Note: The paper explicitly states that because this was a "snapshot" study (looking at one moment in time), we cannot say for sure that AI causes the pain, only that they are tightly linked in this specific pattern. It also notes that the sample was mostly older women in Saudi Arabia, so these results might look different for younger people or men.

Technical Summary: A Network Analysis of Loneliness in Intimate Relationships and AI Attachment Among Married Saudi Couples

Problem Statement
The study addresses a significant fragmentation in current literature regarding the intersection of Artificial Intelligence (AI) attachment and marital loneliness. While existing research has examined marital loneliness within human dyadic dynamics and AI attachment in the context of general social loneliness or trait anxiety, no study has integrated these phenomena within a marital framework. Specifically, there is a lack of understanding regarding how the multidimensional components of marital loneliness (detachment, emotional hurt, guilt) interact with the components of AI attachment (emotional closeness, social substitution, moral appreciation). Furthermore, the psychometric instruments used to measure these constructs—the AI Attachment Scale (AIAS) and the Loneliness in Intimate Relationships Scale (LIRS)—had not previously been validated in Arabic-speaking samples. The authors posit that without a network-based analytical approach, the reciprocal interplay and structural architecture between these variables remain obscured, leaving clinicians and researchers without a coherent map of how AI use and marital distress shape one another.

Methodology
The study employed a cross-sectional correlational design with a sample of 306 married individuals residing in Saudi Arabia. The demographic profile was predominantly female (75.5%), with a majority aged 35 or older (51.0%) and in long-term marriages (over 8 years for nearly two-thirds of the sample).

Instruments and Validation:

AI Attachment Scale (AIAS): A 15-item scale measuring three dimensions: Emotional Closeness, Social Substitution, and Moral Appreciation. Confirmatory Factor Analysis (CFA) confirmed a three-factor correlated model with acceptable-to-good fit indices (CFI = .96, TLI = .95, RMSEA = .057). The scale demonstrated high internal consistency (Cronbach's $\alpha$ = .90).
Loneliness in Intimate Relationships Scale (LIRS): A 14-item scale measuring three dimensions: Detachment, Emotional Hurt, and Guilt. CFA supported a three-factor structure, though with marginal-to-adequate fit (CFI = .91, TLI = .89, RMSEA = .077). Internal consistency was excellent (Cronbach's $\alpha$ = .92).

Analytical Approach:
Following the computation of descriptive statistics and Pearson correlations, the authors utilized Exploratory Graph Analysis (EGA) with a graphical LASSO (Least Absolute Shrinkage and Selection Operator) and EBIC model selection to estimate a sparse, undirected network of partial correlations. The network structure was identified using the walktrap algorithm. Node influence was assessed via eight centrality indices: Degree, Strength, Betweenness, Closeness, Eigenvector, Hub, Authority, and Expected Influence. Stability was evaluated using a parametric bootstrap (500 iterations). All analyses were conducted in R version 4.5.0.

Key Results

Network Structure: EGA identified a two-community structure that aligned precisely with the theoretical constructs: one community comprising the AI Attachment subscales and the other comprising the Loneliness subscales. The network density was 0.67, with all edges being positive.
Centrality Findings: Emotional Hurt emerged as the most central node across all eight centrality indices (Strength = 0.97, Eigenvector = 1.00). It functioned as the "psychological engine" of the network, acting as a critical bridge between guilt, detachment, and compensatory AI use.
Cross-Community Connections: Connections between the AI and Loneliness communities were generally weak. However, AI Social Substitution showed the strongest links to marital constructs, specifically connecting to Marital Guilt (edge weight = 0.099) and Emotional Hurt (0.077).
Peripheral Nodes: AI Moral Appreciation was found to be notably peripheral (Strength = 0.461), indicating it is largely disconnected from both marital distress and other forms of AI engagement.
Global Metrics: The network exhibited 100% positive edges, suggesting that higher scores in AI attachment and marital loneliness covary in a consistent, unidirectional manner without trade-offs. The average path length was low (0.140), indicating that activation or change in one node (particularly Emotional Hurt) could rapidly propagate across the network.

Key Contributions

First Network Analysis: This study provides the first application of network analysis to model the interplay between AI attachment and marital loneliness, moving beyond simple correlation to map the structural relationships between specific symptoms and behaviors.
Psychometric Validation: The research successfully validated the AIAS and LIRS for use in Arabic-speaking samples, establishing their factor structures and reliability within a Saudi cultural context.
Identification of the Core Node: The study identifies "Emotional Hurt" not merely as a symptom but as the organizing principle of the marital loneliness network, suggesting it is the primary driver of the system's dynamics.
Clarification of AI Roles: The findings differentiate between AI dimensions, revealing that "Social Substitution" acts as a bridge to marital distress, whereas "Moral Appreciation" remains peripheral to the experience of loneliness.

Significance and Claims
The authors claim that their findings challenge the assumption that AI attachment serves as a relief mechanism for marital loneliness. Instead, the network analysis suggests a paradoxical dynamic where turning to AI for social substitution reinforces and deepens marital pain (guilt and emotional hurt) rather than alleviating it.

The study argues that Emotional Hurt is the linchpin of the network. Consequently, the authors suggest that clinical interventions should prioritize the repair of emotional hurt within the marriage rather than directly targeting AI usage. They posit that addressing this core node may destabilize the entire architecture of distress, potentially reducing guilt, detachment, and the reliance on AI as a compensatory mechanism. The paper concludes that while humans can form attachments to machines, in the context of marital loneliness, such attachments may undermine human flourishing by reinforcing a self-amplifying cycle of relational distress.

Limitations
The authors acknowledge that the cross-sectional design precludes causal inferences regarding the directionality of relationships. The sample's demographic skew (predominantly female, older, and long-term married) limits generalizability to younger or male populations. Additionally, the reliance on self-report measures and the absence of negative edges in the network are noted as constraints.