AI model suggests that dreams shape daily spirituality over time

A new study reports that dreams featuring supernatural beings or events can influence a person’s feelings of closeness to God in their waking life. This effect is not always immediate, with the research showing that dream experiences can impact daily spirituality with a delay of several days. The findings were published in the journal Frontiers in Psychology.

Researchers conducted this study to explore a long-standing but seldom scientifically tested idea: that religious and spiritual concepts, particularly ideas about supernatural agents like gods or spirits, might have origins or be maintained through dream experiences. Billions of people worldwide identify as religious, and their conception of God can significantly affect their mental health, well-being, and social behaviors.

For instance, viewing God as kind and loving is associated with positive outcomes, while viewing God as harsh and authoritarian can be linked to anxiety and social mistrust. Given these connections, understanding the cognitive mechanisms that shape and update a person’s image of God is an important area of psychological science.

The brain’s activity during the dreaming phase of sleep, known as rapid eye movement sleep, creates a unique mental state. It is characterized by intense visual imagery, strong emotions, and a hyper-associative mode of thinking, all while the brain’s capacity for logical reflection is reduced. This state is thought to be well-suited for simulating extraordinary encounters, including interactions with supernatural beings. The study aimed to use modern longitudinal methods, tracking people over time, to see if a measurable connection exists between the content of dreams and a person’s daily spiritual life.

To investigate this, the researchers conducted an intensive two-week study with 124 healthy adults. Participants completed surveys every evening and every morning from their own homes. These surveys tracked their daily moods, social interactions, and spiritual feelings, including a daily rating of their felt closeness to God and their conception of God as authoritarian. Each morning, participants also reported any dreams they could recall. These dream reports were then analyzed for specific characteristics.

Free daily newsletter

The team used a combination of human rating and machine learning to analyze the dream reports. Three trained assistants scored each dream for “dreamer agency,” which is the degree to which the dreamer felt in control of events within the dream. A machine learning classifier was trained to identify dreams containing supernatural content. Participants also rated their own dreams on scales like pleasant or unpleasant, and strange or familiar. In addition to the surveys, approximately half of the participants wore a specialized headband with electroencephalogram sensors to objectively measure their sleep stages, including the percentage of time spent in rapid eye movement sleep.

The initial analysis focused on the characteristics of supernatural dreams. The researchers found that dreams classified as containing supernatural content were consistently rated by the dreamers themselves as more bizarre, strange, and scary compared to non-supernatural dreams. A key finding was that in these supernatural dreams, the dreamer’s sense of control, or agency, was significantly lower. This suggests a link between feeling a loss of personal power in a dream and the appearance of powerful, non-human characters or events.

Next, the researchers used statistical models to connect these dream variables to the participants’ daily spirituality ratings. They found that the emotional tone of a dream, its affect, had a direct impact on the following day. More positive dreams were associated with higher next-day ratings of closeness to God. The amount of rapid eye movement sleep a person got also predicted a greater feeling of closeness to God and a stronger conception of God as authoritarian.

The analysis also revealed a delayed relationship between dreams and spirituality. The models showed that a person’s dreamer agency and dream affect from four nights prior were predictive of their current day’s feeling of closeness to God. Overt supernatural content in a dream did not predict next-day changes, but it did have a negative association with closeness to God at a four-day lag. This suggests that the cognitive and emotional impact of certain dream experiences may take several days to fully integrate and influence a person’s waking spiritual perceptions.

To further test and understand this delayed effect, the researchers conducted a second phase of the study using a more advanced computational technique. Standard statistical models can struggle to capture complex, non-linear relationships that unfold over time. To address this, the team built a temporal sequential artificial neural network, a type of computer model designed specifically to learn from sequences of data over time. They fed this network the same data from the first study to see if it could more accurately predict the daily spirituality ratings.

The artificial neural network significantly outperformed the original statistical model in predicting daily feelings of closeness to God. The researchers then used a technique called saliency map analysis to peer inside the “black box” of the network and see which pieces of information it considered most important for its predictions. This analysis confirmed the findings from the first study: the network placed a greater emphasis on dream-related data from three and four days in the past than it did on data from one or two days in the past.

In a final step, the team tested several versions of the neural network, each with a different time window. They compared models that could only see data from one day ago, two days ago, three days ago, and four days ago.

The results showed that the models with three-day and four-day lags were significantly more accurate than the models with shorter one-day or two-day lags. This provided strong computational support for the idea that the influence of dream content on waking spirituality is a process that unfolds over several days. The network was not, however, able to successfully model the changes in the authoritarian God concept, which the researchers suggest may be due to limitations in the amount of data available.

The researchers propose that these findings support the hypothesis that dreams serve as a space where concepts of supernatural agents are generated and regulated. The process may begin with a reduction in the dreamer’s sense of agency during a dream. This cognitive state could create an opening for the mind to simulate other characters who possess the agency the dreamer lacks, sometimes taking on extraordinary or supernatural powers. These emotionally charged simulations are not always processed overnight. Instead, the mind may take several days to integrate the experience, which then influences waking beliefs and feelings, such as the felt relational closeness to God.

The study has some limitations. The machine learning method used to classify dreams as supernatural is not perfect and may misidentify some content. The study sample was recruited in part using advertisements for people with frequent nightmares, which may have resulted in a group with higher-than-average negative dream affect. The use of at-home sleep monitoring headbands, while practical, can be less precise than laboratory-grade equipment and is subject to user error. Finally, the artificial neural network analysis was performed on a relatively small dataset, which may limit the generalizability of its findings, particularly for outcomes it was unable to model effectively.

Future research could build on these findings by using different methods to capture the ongoing, dynamic shifts in emotion and belief that follow a significant dream. Further work with larger datasets could also help to confirm the results of the neural network and explore its inability to model certain spiritual concepts. The study represents a notable step in applying modern longitudinal and computational methods to explore the complex relationship between the inner world of our dreams and the foundational beliefs that shape our waking lives.

The study, “Dream content influences daily spirituality ,” was authored by John Balch, George Hodulik, Rachel Raider, Aidan David, Chanel Reed, Wesley J. Wildman, David Rohr, and Patrick McNamara.