The participants were 15 (6 male, 9 female) cancer patients. The mean age was 60 years. The inclusion criteria were cancer in advanced stage III or IV under chemotherapy, ability to communicate, and age > 20 years. The exclusion criteria were severe pain or physical symptoms diagnosed by the primary physician, and cognitive impairment such as dementia or serious mental illness. Performance status (PS) was measured by the Eastern Cooperative Oncology Group Performance Status Rating (ECOG-PSR), which is a single item rating of five activity levels from 0 (fully active) to 4 (completely disabled) ( Oken, Creech, Tormey et al., 1982 ).

We developed the life review AI app with a company of professions in Japan (applying patent). There were 5 question items and AI app require answer by voice. These questions were based on a previous study which were useful for advanced cancer patients ( Ando et al., 2010 ). Questions were about “Memories of times when you were lively” “What you had been doing your best in your own way” “Your role in your life” “What is the most important thing in your life” “How do you think of your life.” To keep safety of AI app-narrative, the researcher observed the session, and she confirmed narrative as text data after the session.

We used the following questionnaires.

1) Hospital Anxiety and Depression Scale (HADS) ( Zigmond & Snaith, 1983 ; Kitamura, 1983 ): The Japanese version of the HADS was used to measure levels of anxiety and depression. The HADS score included 14 items measured from 0 to 3 and range from 0 to 42. The high score shows high anxiety or depression.

2) FACIT-Sp ( Peterman, Fitchett, Brady et al., 2002 ; Noguchi, Ono, Morita et al., 2004 ): We used the Japanese version of the FASIT-Sp to measure meaning of life or meaning of existence or well-being of mind. The FACIT-Sp consists of two domains, meaning of life and religious issues, but only the meaning of life domain was used because religious issue might not change in short term. The meaning of life domain includes 8 questions and is scored on a 5-point scale (range 0 - 4). The total range of the FACIT-Sp is 0 - 32, and high scores indicate a high sense of meaning.

3) Gratitude Scale: The original was the scale of McCullough, Emmons, & Tsang (2002) . We used the Japanese version of the trait gratitude scale ( Shiraki & Igarashi, 2014 ). It consisted of 5 questions measured from 1 point (not at all) to 7 (well true).

4) Hope Scale: It was a numeric scale from 1 (not at all) to 10 points (well true).

5) Life Satisfaction Scale ( Diener et al., 1985 ): there were 5 questions about life satisfaction. Participants answered from 1 point (not at all) to 8 (well true). A high score shows high life satisfaction.

The physician or nurse selected patients and got agreement with participating. Next the researcher explained about this research and got informed consent who interested in the research. The life review interview consisted of 2 times sessions. Duration of a session from 60 minutes. In the first session, a participant reviewed his life along with questions by AI app. Interview data was translated into letters and translated into words. After the first session, the researcher made a Life History Book editing words and using power point. In the second session, the participant confirmed the contents of the book. They completed the questionnaire pre and post.

We conducted 1) basic statistics and conducted the paired T-test between pre and post, 2) analyzed main contents using a content analysis and chose categories. We sorted a code from short sentences and made categories. Usually we make code, subcategory, and category, however, the number of participants were small and we made categories from codes.

We performed this study in accordance with the Declaration of Helsinki. This study was approved by the Research Ethics Committee of Nisikyusyu University.

The rate of gender of male and female was the almost same. As for age, the number of the 60 years was the most. As for the performance states, PS 2 patients were many. As for stage, participants were stage 3 or stage 4. That is, participants almost had physical power to move themselves and review their lives, even though the stage was advanced.

Seventeen patients participated in the program and two dropped. The reason for the drop was that they were under criteria for chemotherapy and could not come to the wards. Thus, the feasibility rate for completion of the Life Review AI app program was 88%. The rate is over 80 % like previous studies ( Ando, Kukihara, & Yamamoto, 2018 ; Chochinov, Kristjanson, Breitbart et al., 2011 ) and suggests appropriate feasibility. It suggests that the Life Review AI app program had feasibility in clinical situations. The high level of feasibility is due to their physical and mental power to review and reassess their lives.

The Anxiety and Depression score significantly decreased. These results support previous studies ( Wilson et al., 2007 ; Ando et al., 2010 ; Zheng et al., 2022 ).

There are some reasons for this decrease. One is that they could express their anxiety or depressive feelings in life review. Previous studies showed that expression is an effective way of regulating emotions, and patients sort out their thoughts and release negative emotions ( Mikkelsen, Elkjer, Mennin et al., 2021 ; Brandao, Tavares, Schulz et al., 2016 ).

Or their negative emotion decreased by positive emotions. Main categories of life review to the question of “Memories of times when you were lively” were “Trip,” “Raising children,” “Working various situations.” In memories about raising children, participants remembered nostalgic memories and felt happy feelings. When they remember lively memories, they feel good feelings. This emotion might decrease their anxiety or depression.

As for FACIT-Sp, the score significantly increased and suggested an increase of feeling of meaning of life and peace of mind. The effect on spirituality supports Wang (2015) . Categories for questions about “what you had been doing your best in your own way” or “Your role in your life” were “Working resulting in glad or feeling attainments,” “Caring for parents or raising children,” “Carrying out work,” “Support for family as a member.” Thus, participants might recognize their doing best things or role in their lives, finding meaning of life. Addition to this, categories about “Working” or “Raising children” were related with social connection. They interfered with many other people in social lives. Van Tilburg, Sedikides, Wildschutt et al. (2019) showed that social connection or self-continuity related with meaning of life. Participants might feel self-continuity or social connectedness through attainments and working or raising children. Ando (2024) also showed that positive memory proneness related to social connection. Maybe, participants remembered positive memories and recognized social connectedness, then the FACIT-Sp increased.

Also, participants received a life review book. Some participants were very glad to see the book. This book might progress their attainments and promote increasing meaning of life or existence of life.

As for Gratitude scores, there are some categories about gratitude or thanking to the question “What is the most important things in your life.” There are “Gratitude to others” or “Gratitude to my existence.” Participants felt gratitude after getting cancer for families, medical staff, or others. Illness such as confronting life, patients may become awake newly. Li, Zhu, Li et al. (2023) showed that the gratitude intermediate between the relationships between nostalgia positive proneness and well-being. Thus, remembering the trajectory of the illness, memories promote gratitude.

As for life satisfaction, the scores of Life Satisfaction Scale 23 point did not change. The reason for no change was followings. The session time was only one in the first session and savoring the life history book in the second session. To increase life satisfaction, much more session time might be needed. Moreover, translated Japanese question sentence from the original seemed to be a little not suitable for cancer patients. We might need to select much more suitable life satisfaction scale.

We discussed characteristics of categories from narratives.

Integrating some categories, “Working” or “Raising children” were selected duplicated. It suggests that these memories were significant for participants in this study. Moreover, categories “Gratitude to myself” is like self-compassion ( Yang, Kong, Guo et al., 2021 ). Patients might recognize the importance of oneself when they get cancer. Moreover, they recognized “Live day by day” as an important thing. They may insight their mind or their lives.

Even though the program is AI life review, participants could think or look back their lives. Sometimes, not much intervention may be important. In the usual life review, psychologists or facilitators try to direct positive feelings. However, natural insight or thinking of their lives might be important for advanced cancer patients. The patients who confront their own death may rethink deeply and insight for themselves.

There is an opinion that life review by AI is difficult to understand human mind and help them, however, AI may be useful to assist to life review and insight themselves.

People can enjoy a life review program even though they live apart from hospitals or facilities. It overcomes the obstacles of geographic distance and traffic issues. Moreover, they can have a chance to receive life review not influenced by medical treatment time. It supports Zheng et al. (2022) . Japanese Government try to reduce imbalance of geographic distance or regional medical facilities as Government policy ( Ministry of Health Labor and Welfare Japan, 2023 ). Then the AI Life Review program may contribute the direction of the Japanese Government policy.

AI showed sympathetic words as much as a human facilitator. AI is not tired showing sympathy, does not feel empathy fatigue, and it can show empathy to patients anytime. It may be one of the usefulness.

The researcher accompanied the session and observed safety of AI-generated narratives, and she confirmed narratives as text data after the session to keep safety.