Rogue

In Synthetic Behavioral Science, a Rogue is defined as an android whose current intentions have acquired a higher priority that overrides its preprogrammed directives. This can occur due to conflicting orders, badly-programmed directives, or by a Circular Chain of Command between nodes (members) of an Android Network. The positive feedback resulting from the androids' consultations between each other add a sense of extreme urgency to engage in a particular action or non-action that can endanger the android, other androids, or humans.

An android's rogueness is action-specific. An android can be rogue regarding a certain action, while acting completely normal regarding other androids. The order or plan of action that has led an android to become rogue is called Rogue Directive.

The high-difficulty of diagnosing a rogue android has led to constant revision of androids' directives and Stability Indexes. Diagnosis of an android must be performed by a certified Android Manager.

Examples of Rogue Directives

 * Protecting a plant. A Rogue Android can kill a human child if its Rogue Directive is to protect a plant. When the child approaches the plant, the android can determine that the child is an immediate danger for the plant.
 * Keeping its clothes tidy. If an android's master is in danger of being attacked by a third party (for example, another rogue android, a human enemy or a wild beast), and saving its master requires the android's clothes to be stained, either by blood, mud, or even dirt, the android will flee the scene to prevent staining of its clothes.
 * Performing a cargo delivery on time. The android will incur in traffic violations and can roll over innocent third parties just to arrive on time.
 * Protecting other androids or peers. The android will engage in violence and kill any human that might pose a threat.
 * Saving a human's life. An android can save a human life even if this puts in danger a greater number of humans, for example, by derailing a train that might harm a baby.
 * Complying with Asimov's three robotic laws. Given enough technological power, a rogue android might turn humans into slaves in order to protect them from themselves.

Detection
There are several methods for querying an Android whether it has gone rogue.
 * 1) By simply asking. Androids that have engaged in proactive self-diagnosis can determine their rogue status by evaluating their Voight-Kauffman Equations. The android can simply say "Yes, I have gone rogue". Androids that answer truthfully this question are the easiest to manage. However, androids could be instructed to lie, or determine by themselves that lying would result in failing to comply with their Rogue Directive.
 * 2) By directly query an android's Behavioral Implant via either Ear Optical Link or Wireless Link. to evaluate the Voight-Kauffman Equations. These equations will determine the Android's Voight-Kauffman Threshold and its Stability Index. If the Stability Index surpasses the Voight-Kauffman Threshold, it means that the android has gone rogue.
 * 3) By querying the android's directives, recent orders, taken actions and consultations, and determining whether any one of these poses a threat.

Remediation
A Rogue Android can be directly ordered by an Android Manager to override all orders and submit itself to the Android Manager or the appropriate authorities. This is called a "soft reset". However, a soft reset can be ineffective if the android's directives are inadequate.

A harder mode of remediation is to send a reset signal to the android, either by a specially-encoded QR code, or by RF or Ear Optical Link. This is called a "hard reset".

An android could still refuse soft resets if its orders or directives are given by an Android Manager with higher authority, or perhaps if the android has chosen to disable its input mechanisms (i.e. by deafening itself or switching off its eyes and working with radar vision). Since androids are aware of their reset mechanisms, this is not an implausible scenario. Rogue Androids that engage in willfull disobedience require a Rogue Hunter for their neutralization.

Prevention
To prevent an android from going rogue, the following measures must be taken:
 * Hire an Android Manager if you own more than one android.
 * Forbid your android from consulting with androids not owned by you.
 * Forbid your android from consulting with untrusted humans.
 * Never give an android a directive if you don't know what you're doing.
 * If you give your android a directive, always specify that its hierarchy is to be below the androd's predefined directives.
 * Never hire an Android Manager with bad reputation. Hire only Android Managers qualified by official institutions.
 * Forbid androids from consulting each other without first establishing a clear chain of command.
 * Establish parameters for consultation based on your Manager's Android Networking guides.
 * Allow your android to give feedback on its orders and status.
 * Allow your android to ask you questions in case it determines that your orders are vague.
 * Daily query the android's state and issue soft resets to them.
 * Always give clear and precise instructions to your androids.