**5. The balance between human and machine**

When trying to find an adequate balance between human and machine collaboration in the OR the subject of autonomy is a natural starting point. Surgical teams today are comprised of highly specialized professionals that need to work in perfect synchrony for surgical procedures to run smoothly. The surgeon, as the leader, must find balance between managing everything going on with a high degree of control, whilst still allowing for the autonomy and independence of each team member. Most surgeons are authoritative leaders within these teams, meaning that they retain control while still empowering the freedom of self-management where each member can be engaged, motivated and focused on their personal tasks at hand [52]. Although the surgeon is ultimately responsible, he or she will not intervene in a nurse's needle or instrument counts, or check whether the anesthesia machine is properly working. Surgeons authorize themselves to relinquish this direct control because via a strong culture, values and guidelines they ultimately continue to provide the critical oversight and supervision for effective risk-management [52].

Besides team management, the surgeon may be liable for equipment malfunctions, therefore there is a certain underlying hesitancy in giving autonomy to machines. A 2013 systematic review of surgical technology and operating room safety failures found that up to 24% of errors within the OR are due to equipment malfunction [53]. This has not, however, stopped us from relinquishing control in certain parts of the surgery and delegating it to tools which we cannot always fully control. Advanced hemostatic devices like Ligasure™ for example, automatically adjusts and discontinues the delivery of energy based on its own calculations without any surgeon input. Similarly, the Signia™ Stapling System has Adaptive Firing™ technology that automatically and autonomously makes adjustments depending on the tissue conditions it senses [54, 55]. So, while there is hesitancy from the surgeon side for adopting new autonomous devices, if the surgeon is able to see the benefits as with the Ligasure™ and Signia™ systems, these types of tools can in fact break the barrier of more advanced machines into daily OR practice.

#### **5.1 Machine autonomy in other fields and how they can relate to the OR**

Whether we are aware of it or not, AI is already affecting the world and making our everyday lives easier. It is there every time we search for something online. It automatically recognizes us in pictures we take, it recommends new music, food or products we will like. AI helps us hear what is written and read what is spoken. It protects us from credit card fraud and helps us make smarter investments. At home it manages our thermostat and decides when and where to vacuum clean the floors.

Moreover, machines are already responsible for millions of human lives on a daily basis, albeit indirectly. The oldest and most famous example is probably the autopilot in airplanes; multiple studies have shown that in 95% of commercial flights, pilots spend less than 440 seconds manually flying the plane [56, 57]. Other examples include the automation of emergency medical service dispatchers and the automation of trains and metros, where nearly a quarter of the world's metro systems have at least one fully automated line in operation [58–60].

The advancements of automation in settings where human lives are at stake have pushed society to further debate the autonomy versus control issue. Depending on the field, different scales have been proposed to define levels of automation and

autonomy. These scales have been important as they help define the capabilities and limitations of a system's autonomous features and establishing expectations around the operator's behavior and responsibilities. In addition, they have helped build trust and reduce anxiety around autonomous machines, while ensuring that legal and ethical concerns are considered as technologies continue to develop.

The most prominent autonomy scales revolve around the automotive and aviation industries. The main difference between the two is that the automotive scale encompasses all the systems in a car as a single unit and the vehicle is labeled depending on its capability as a whole [61]. While in the aviation scale, each system in an airplane receives a level of automation independent from other available autonomous systems on the same plane [62]. It is important to note that the scales defining the levels of autonomy in cars, trains, and planes all have basic similarities which are adapted to each specific industry. These adaptations are dependent on the level of complexity of each industry, and the training of the average operator. All the scales, across the various industries begin at level 0 where there is no automation at all, gradually increasing to level 5 (or the maximum of 4 in trains [63]) where there is full machine autonomy without the need for human input at all.

In the field of surgery, the question of how to define the levels of autonomy in systems within the OR has already begun, and although surgical systems are not yet as advanced as other industries', it is important to have a standardized language when referencing this subject. Yang GZ et al.'s proposal for defining the levels of autonomy for medical robotics [64] has been extremely effective in catalyzing the debate of defining the levels of autonomy in surgery. This scale is loosely based on the automotive levels of autonomy as it grades a robotic system as a whole depending on all of its capabilities.

The scale is composed of 6 levels (0–5) as follows:


Others have built upon this scale, using similar classification methods for surgical robot autonomy [65, 66]. Current technology in robotic surgery is only at Level 0, but when the objectives of the research projects described above are met, we might reach level 1 and 2.

As surgeons, our experience in the OR environment is more comparable to flying a sophisticated airplane than driving a car. A surgeon's professional responsibility is
