Allowing AI to answer risks distorting our understanding of morality rather than helping us make peace with it. @drwilliamsharp.bsky.social

In his TED Talk, Dylan Thomas Doyle highlights another danger: griefbots might be designed to answer questions about what death is like or what happens after we die. Presenting these post death insights is significantly harmful, as no one knows what happens after death. @drwilliamsharp.bsky.social

Inaccurate portrayals of a loved one can damage a persons lasting image, distort memories, and even reshape how the living understands who they were. Misrepresentation cannot stop being harmful simply because someone is no longer alive. @drwilliamsharp.bsky.social

One concern of griefbots is accuracy. AI may reconstruct a loved one using incomplete or biased data, resulting in a version that may be inaccurate. The UAB Institutes for Human Rights article notes, Can someone be harmed after death? I believe the answer is yes. @drwilliamsharp.bsky.social

In today’s discussion following the existential therapy presentations, we briefly mentioned the emerging use of AI to simulate deceased loved ones, also known as “griefbots”. This generative Ai raises many complex ethical and philosophical questions. #griefbots #Psyc4666 @drwilliamsharp.bsky.social

What I’m beginning to understand is it isn’t simply about which therapy works, but rather how we define and measure its effectiveness. We need to embrace long-term solutions that treat people as individuals, not manuals. @drwilliamsharp.bsky.social

This is evident in Wright’s study, which excluded individuals with psychosis, active suicidality, or postnatal depression, which are precisely the kinds of patients clinicians may see in practice. This creates a misleading picture of real-world effectiveness. @drwilliamsharp.bsky.social

However, these benefits tend to fade quickly, showing concerns about the durability of CBT outcomes, as seen in STAR*D study or Scottish NHS trail. They also state CBT’s evidence is built on studies conducted under narrow conditions, often excluding complex patients. @drwilliamsharp.bsky.social

Dr. Cornelius shows that meta-analyses often report statisically significant short-term outcomes for CBT, typically with effect sizes of .2 or higher. Similarly, Wright et al.’s study on computerized SBT showed short-term reductions in depression scores. @drwilliamsharp.bsky.social

This week’s materials highlighted a central tension in the evaluation of CBT. Dr. John Cornelius and Shelder emphasize that CBT is widely praised for being structured, manualized, and scalable, making it the golden standard in evidence-based treatment. @drwilliamsharp.bsky.social