Chatbots in cancer applications, advantages and disadvantages: all that glitters is not gold

“Instead, he presented it as a debunking exercise [10].”

“…in the healthcare industry may face substantial challenges navigating regulatory environments; besides, the fast developing field of chatbot software and the absence of industry standards in chatbot applications exacerbate the difficulties involved in regulatory assessment and supervision [82,83].”

“Despite the fact that these AI technologies reduce customer costs, they have high initial development costs and maintenance costs [56].”

“…automated chatbot models within the healthcare industry follow the rules of justice and equity because of the possibility of unfavorable results, because reaching this goal can support fair healthcare outcomes and access for all populations, irrespective of their demographic makeup [79–81].”

“While basic chatbots have existed for decades, contemporary versions often incorporate natural language processing and deep learning [6].”

“ALICE avoids improper responses by using pattern-matching algorithms to retrieve example sentences from predefined output templates [11].”

“While generative neural networks, which are built on algorithms that use deep learning to produce novel answers word by word depending on user input, are often trained on an enormous amount of natural language phrases, chatbots require a large quantity of conversational information to train [60].”

“…have multifarious advantages, with the most important being that they can provide support 24/7 at any time with instant response regardless of space, and also they provide answers automatically, thus achieving heightened lifetime value, robust brand affinity, and higher levels of satisfaction [34].”

“…since chatbots depend on an internet connection to work, users in places with spotty or inconsistent internet connectivity may find it difficult to use chatbot services, yet it is important to take into account the user base’s potential restrictions, particularly with regard to patients [55].”

“Despite their acceptance by doctors, there are still issues with chatbots’ incapacity to understand human emotion and situations when medical intelligence and expertise are needed [87].”

“…in the healthcare industry may face substantial challenges navigating regulatory environments; besides, the fast developing field of chatbot software and the absence of industry standards in chatbot applications exacerbate the difficulties involved in regulatory assessment and supervision [82,83].”

“In contrast, AI chatbots, also known as AI conversational agents, process a natural language using neural networks or algorithms instead of relying on pre-written scripts [7].”

“…Chatsonic, and Bing AI for the top Google search queries pertaining to five common cancers (lung, skin, colorectal, breast, and prostate), and despite having an excellent median DISCERN score, the replies were difficult to grasp and not easily actionable at the collegiate reading level [70]. v3.”

“Conversations with users that are non-linear and need back-and-forth exchanges are challenging for chatbots to handle [58].”

“REVIEW Differences between the guidelines and the chatbot’s output were frequently ascribed to unclear answers, suggesting that care should be used when utilizing chatbots to obtain treatment-related information [69].”

“Recognizable web-based voice-activated assistants such as Microsoft Cortana, Apple Siri, Amazon Alexa, and Google Assistant have become widespread following the popularity of chat apps [12].”

“Importantly, a study on women with breast cancer found that the chatbot’s EORTC INFO25 scores were not less than the doctors’ scores in providing the answers the women needed [54].”

“When humans use and rely on chatbot-generated content tainted with delusions, it is referred to as “botshit” [8].”

“Moreover, in this way, some more efficacious patient self-services can be seen, since chatbots can help users to reach the appropriate data and stay healthy and safe, and they can provide a second and a third opinion on a diagnosis or medication plan, regardless of the user’s location [47].”

“Lastly, human computation is used in human-aided categorization, which offers greater robustness and flexibility but is slower to process more requests [32,33].”

“Things are not always so bad, but miscommunication takes place [74].”

“…automated non-human system and share personal information and very sensitive health matters, and chatbots may not be as trustworthy to some as a genuine person, who can give them tailored advice and respond to their questions instantly—particularly for cancer or other serious health problems [65].”

“Such misdiagnosis can occur due to various reasons: erroneous system diagnosis, wrong estimation from AI, inadequacy of chatbot questions, and inaccurate data provided by the user either due to distress/uneasiness or underestimation/negligence of their symptomatology [67,68].”

“It should also be mentioned that, apart from acceptability, health literacy matters too; users should be capable of reading and comprehending the chatbot’s outputs, yet there are various users with different knowledge, and this is a next-stage limitation of digital health [84,85].”

“Additionally, they may offer emotional support to cancer patients and enhance oncology care [51].”

“…handle private or sensitive client data might lead to security issues, and it is essential to make sure that data are encrypted, stored safely, and shielded from unwanted access, because security lapses may result in serious repercussions, such as harm to one’s reputation and legal status [61,62].”

“…and trigger again a false diagnosis; however, fake results and misdiagnosis are possible not only in digital diagnosis but also in clinical and laboratory diagnosis—the last was recently seen in false COVID-19 test results and even Progressive Multifocal Leukoencephalopathy diagnosis [90–96].”

“Personalization is another challenge of chatbots, yet the difficulty lies in figuring out the best ways to adjust to the user, and it can only be resolved by making several attempts and failing in each unique situation [71–73].”

“…scheduling of appointments integrated with online calendars and automatic reminders of them, diagnoses of rare or tricky illnesses and interpretation of complex diagnostic tests, and a lot of support for the person’s overall health, from physical activity to behavioral and mental health needs [46].”

“It has been said that chatbots in medicine are a medical supplement and not a substitute, because they remain short of the human qualities of empathy, instinct, and prior experience that make healthcare workers human [78].”

“…healthcare, and it is essential for early health issue discovery, treatment plan adherence, patient education, and general patient happiness; however, traditional healthcare communications on strategies have faced both opportunities and challenges with the arrival of the digital age [19].”

“…practical and affordable tools to help improve and increase self-care practices and reduce the side effects of chemotherapy, and they can also act as empowering tools to support nurses in educating women with breast cancer and empower them to take an active role in dealing with their symptoms [53].”

“It should also be mentioned that, apart from acceptability, health literacy matters too; users should be capable of reading and comprehending the chatbot’s outputs, yet there are various users with different knowledge, and this is a next-stage limitation of digital health [84,85].”

“Thus, physicians can more easily control patient demand without incurring additional fees, thanks to this scalability, and they can also maintain low costs while providing top-notch customer service [49].”

“It was discussed in the previous section that the main drawback of chatbots in oncology is that they must undergo rigorous user testing and iterative development before being widely used [52].”

“Nowadays, chatbots are widely used in messaging apps, internal company platforms, politics, toys, data security, education, and healthcare [4,13–18].”

“Particularly in the field of healthcare, to be able to identify specific patterns in big data, developers and suppliers of AI/machine learning-driven medical products need a significant amount, speed, variety, and veracity of medical data [76].”

“Lastly, human computation is used in human-aided categorization, which offers greater robustness and flexibility but is slower to process more requests [32,33].”

“As with all diseases and other reasons to use a chatbot, challenges include data privacy, security, and ethical aspects [88].”

“…on the most important discussions, and in this way, they help in reducing daily stress and anxiety and coping with stressful situations, not to mention also that chatbots can deliver messages that are emotionally comforting and customized to the stressors that users share with them [35,36].”

“The goal of contemporary chatbots is to converse with people in a conversational manner, and they can offer support in a friendly and engaging manner, as well as answer inquiries and make product recommendations [37].”

“…services, this may indicate that numerous organizations have gained access to your private details, while big data and AI can be costly for startups or small businesses that do not currently have the resources to operate efficiently or still need access to this kind of technology [50,66].”

“…fact that some chatbot systems support multilingual text and speech in a fraction of a second, and they are crucial especially in healthcare, where they can support changes in behavior and they can also act like assistants and help with routine tasks and activities in living environments [38,39].”

“…dialogue management that monitors the conservation’s flow and determines suitable responses, the integration layer that links it to external systems and databases to retrieve data or to act, and the analytics and feedback that monitor usage metrics and collect feedback for its improvement [30,31].”

“Importantly, they display visual content that is more descriptive and realistic, and they provide compliance and security for the user’s personal information [41–44].”

“…on data, as well as a lack of time or resources to undertake research while simultaneously managing and caring for human lives; chatbots run a higher risk of producing erroneous or defective solutions since they require constant problem solving when using a network-generated algorithm [63,64].”

“…handle private or sensitive client data might lead to security issues, and it is essential to make sure that data are encrypted, stored safely, and shielded from unwanted access, because security lapses may result in serious repercussions, such as harm to one’s reputation and legal status [61,62].”

“…to overcome these obstacles have been recommended, such as regularly updating the chatbot’s base of knowledge, incorporating machine learning and natural language processing, administering data validation and verification, ensuring data safety, and providing human assistance wherever possible [45].”

“…given that there are several divergent patient types, such as youngsters, elders, those with problems with their senses, those with mental or/and behavioral issues, those who speak English or not, and those coming from other countries, many chatbots provide a better patient engagement [48].”

“…automated chatbot models within the healthcare industry follow the rules of justice and equity because of the possibility of unfavorable results, because reaching this goal can support fair healthcare outcomes and access for all populations, irrespective of their demographic makeup [79–81].”

“Nowadays, chatbots are widely used in messaging apps, internal company platforms, politics, toys, data security, education, and healthcare [4,13–18].”

“…on the most important discussions, and in this way, they help in reducing daily stress and anxiety and coping with stressful situations, not to mention also that chatbots can deliver messages that are emotionally comforting and customized to the stressors that users share with them [35,36].”

“…and trigger again a false diagnosis; however, fake results and misdiagnosis are possible not only in digital diagnosis but also in clinical and laboratory diagnosis—the last was recently seen in false COVID-19 test results and even Progressive Multifocal Leukoencephalopathy diagnosis [90–96].”

“Such misdiagnosis can occur due to various reasons: erroneous system diagnosis, wrong estimation from AI, inadequacy of chatbot questions, and inaccurate data provided by the user either due to distress/uneasiness or underestimation/negligence of their symptomatology [67,68].”

“Additionally, it was previously discussed that privacy concerns are a common obstacle, along with the user’s capability of providing real answers and his/her trust—which is affected by various factors apart from empathy and emotional intelligence [77].”

“Goumas et al. [20] reviewed the use of AI chatbots in cancer support systems, outlining benefits and ethical limitations.”

“However, additional challenges may arise, such as potential harm to the patient–physician relationship and trust, loss of the human touch in patient care, liability issues from over reliance on technology, and risks of misinterpretation [110–112].”

“…cancer patients, the finding of Goumas et al.,, high-lighted critical challenges such as chatbots may produce coherent sounding but erroneous or inaccurate information, a phenomenon known as “hallucinations” because they predict answers instead of understanding the meaning of those responses [11].”

“Examining the user acceptance of chatbots in cancer care is essential for determining how patients, caregivers, and HCPs incorporate these tools into their decision-making and support systems [21].”

“Conversational AI chatbots have demonstrated the potential to provide patients with responses that are as accurate, empathetic, and readable as those provided by healthcare professionals [7].”

“7 10 As in all areas of health care, the accuracy/reliability of the medical information provided is essential for patients with TBIs due to the complex nature of their treatment, rehabilitation, possible long-term consequences, and the high demand for information.”

“12 Although security and privacy in mHealth applications have been discussed brie fl y in various studies, 13 – 15 there remains a noticeable gap in the literature focusing speci fi cally on conversational agents in healthcare.”

“While chatbots may be more easily applied to diagnostics, treatment, and monitoring [26], progress on interpersonal-type communications is being made, despite challenges noted on the algorithmic development side [45].”

“G. Goumas et al. (2024): Goumas and colleagues explore chatbot applications in cancer care, weighing advantages such as accessibility against challenges like limited empathy [14].”

“None of the recent reviews of digital technology in oncology shed light on this (e.g., (66–70).”

“This finding indicated the need for a more nuanced approach in implementing the study.”

“Previous studies demonstrated that chatbots were effective in medication adherence, symptom monitoring, and patient empowerment in diverse cancer settings [2].”

“Tools such as Florence, Medi Reminder, Conferbot, and Sensely provide medication reminders, guide symptom self-management, and offer psychosocial support [163].”

“Compared with traditional health education approaches, chatbots have many advantages, including the ability to simplify specialized medical knowledge, low costs, and instant response for easy public access [22, 23].”

“However, the term “breaking bad news” is still widely used in clinical practice and studies, especially when referring to the initial disclosure of a serious medical condition [6].”

“[28] “ChemoFreeBot” was found effective in improving self-care behavior and reduce chemotherapy side-effects in breast cancer patients because of its personalized approach.”

“Conversational agents are increasingly utilized in health care [49], and recent advances in large language models have also enabled an increasing use of conversational agents to support personal sleep wellness, for instance, by delivering CBTi–based interventions [50].”

“…enable more natural and personalized engagement, but raise concerns about reliability, transparency, and regulatory compliance, as they process language via neural networks rather than scripts and may generate coherent yet incorrect or fabricated information (Goumas et al., 2024; Oh et al., 2021).”