{
  "abstract": "Clinical artificial intelligence (AI) is proliferating rapidly. Across specialties, new diagnostic aids, predictive risk models and decision support systems are emerging at an accelerating pace, most developed and tested within research settings.1 2 In this perspective, we argue that as their number and complexity grow, they may exceed what the average clinician can reasonably remain apprised of and effectively use—a concern with emerging evidentiary support.3 Each model may carry distinct input structures, interpretation requirements and failure modes, creating a landscape that is increasingly difficult to navigate. For clinicians, the challenge may soon shift from scarcity of reliable AI tools to the practical impossibility of maintaining informed awareness of available tools, appropriate use contexts and their relative strengths and limitations.",
  "authors": [
    {
      "affiliations": [
        "Brigham and Women’s Hospital, Boston, Massachusetts, USA"
      ],
      "name": "Arjun Mahajan"
    },
    {
      "affiliations": [
        "Brigham and Women’s Hospital, Boston, Massachusetts, USA"
      ],
      "name": "Avery H LaChance"
    },
    {
      "affiliations": [
        "Division of General Internal Medicine and Primary Care, Brigham and Women’s Hospital, Boston, Massachusetts, USA"
      ],
      "name": "David W Bates"
    }
  ],
  "title": "Towards a framework for implementing artificial intelligence in clinical medicine",
  "uid": "95716173-e61c-56b7-9b96-bc77bd53e444"
}