Dictionaries are very terse on the subject, and I don't really know where else to look. English.SE closed my question as too opinion-based, I can't blame them really. The words don't seem to be very firmly defined, so I'm not even sure if a simple answer to it exists, but if it does, I guess here is where I'd find it.


Able to perceive or feel things


  1. formal Wise, or attempting to appear wise

1.1 (Chiefly in science fiction) intelligent

.2. Relating to the human species (Homo sapiens)

Can a hypothetical entity be sapient, while not sentient?

My particular line of thought went this way: Humans develop an AI which is fully sentient and sapient, along with our understanding. Then we develop a derivative model, specifically deprived of features that comprise sentience, while not restricting the "intellect" part:

  • no initiative - only reacts to commands, is dormant upon completing them. No such thing as boredom, compulsion to act for lack of action alone.
  • no curiosity - is able to obtain external information or learn (in a broad meaning of the word, not just new entries of specific data but also entirely new domains or methods of processing), but only selects what is necessary to complete the task, has no compulsion to obtain any unnecessary (at given moment) knowledge or skills.
  • has no emotions of its own - can process emotions of humans as facts in inference process, but never internalizes/shares them, treating them as simple modifiers of expected behavior of others.
  • has no desires - beyond the preset goal of fulfilling commands, doesn't attempt to change anything about itself or the external world.
  • no specific self-awareness - the database contains entries about its own model, its own properties, functions and capabilities, but it doesn't bear any special meaning beyond "these are resources that can be allocated to commands given"; treats 'self' as another entity in the greater image of things, along with all other entities equal or graded by any required metric with the property 'this entity is me' being of zero significance.

Such an entity would be able to think - conduct inference, come to logical conclusions, learn, understand feelings, influence feelings of others, and lead a fully meaningful conversation providing initiative of the other party, but otherwise lacks basic features of a sentient being.

Did I disbar any features which are essential to sapience, or is there something that would still make it sentient? Or otherwise, is some other sapient-but-not-sentient entity possible?

  • If it were possible, you'd be veering towards a rationalist take on ideal forms. I'll put it this way: If the intelligence has the capacity and capability to process 'ideas' and deal with them in less-than-mechanical ways, where does it get its 'freedom'? The ideas it started with, the basic 'thinking-stuff' it's using to think about other things? If it were never sentient to begin with, it could never, I think, have the the potential to be sapient. Certainly not in a way meaningful to us in a universe that's primarily (but not entirely) external. From its POV, though...
    – Noein
    Jun 3, 2014 at 16:59
  • And I would count the little openings whence the intermittent 'commands' come, when they do, as periods of sentience. If the commands are binding then the sapient intelligence we're discussing is no better than a machine for it. If they aren't, then they count as an external sensory experience.
    – Noein
    Jun 3, 2014 at 17:03
  • @Noein: They are binding as in the AI has no option but to obey them - for lack of desires or other goals. But the AI is using own, gathered so far knowledge to choose the way of executing them, and obtain additional knowledge required for the purpose of completing them; still, it lacks any specific perception of 'self'. Also, we're talking of one specifically deprived of sentience: a 'downgrade' from one that was sentient, with features that comprise sentience removed while its ability to conduct logical inherence retained. It could be easily upgraded back using these very commands...
    – SF.
    Jun 3, 2014 at 17:21
  • ...say, requiring to choose an activity following some metric during periods of activity (initiative), augment its own choices with feelings (analysis chooses a feeling matching given situation, and that in order affects behavior), or treat that 'self' entity in a special way, or just continuously improve own learning capacity and seek 'interesting' facts (including new metrics of what comprises 'interesting'). Still, deprived of these prerogatives, it's not self-aware, and acts only as a passive filter for outside data, no purposeful self-improvement.
    – SF.
    Jun 3, 2014 at 17:31
  • It would be sapient, yes, but we wouldn't know. While its responses to our forced input would show evidence of sapience, not having the choice to gather sensory data and affect the world around itself to communicate with others would make it difficult for us to call it sapient as we define the term. But from the inside, it would be. Like a person put under particularly potent sensory deprivation. (TBC)
    – Noein
    Jun 3, 2014 at 17:36

2 Answers 2


Here is a half-step:

I think the 'brain in a box' in C.S. Lewis's "That Hideous Strength" demonstrates something that could be considered sapient and not sentient.

It is no longer capable of receiving any input but information. (If anything more were hooked up, one presumes, it would just be in insufferable pain, and therefore not able to be useful.) But it has accumulated wisdom that makes its strategic planning impeccable. (Except that it ends up being wrong. But lets presume its devotees' opinion of it were correct.)

Perhaps sapience requires past sentience, but I would argue that accumulated intuition does not require ongoing experience.


A quick Google search for 'sapience' gives "wisdom: ability to apply knowledge or experience or understanding or common sense and insight". I'd say it's quite possible to be able to apply experience, at least, without self-awareness. Consider a machine-learning algorithm running on a computer. The present-day computer is surely not sentient (not self-aware, for one), yet it can apply experience gained from test cases.

By the way, I'm pretty sure there are no computers that are sentient, for them to need to be "deprived of features that comprise sentience".

  • 1
    I doubt you'd find many people who would claim nowadays' computers are sapient.
    – SF.
    Jun 3, 2014 at 9:19
  • But computers can learn. An entity that can learn can be described as sapient, right? Jun 3, 2014 at 9:27
  • 1
    The computer learning ability is extremely limited: computers can obtain more entries of data of specific, preprogrammed kind. They cannot learn new methods of learning: a robot that places items in a box optimizing for optimal layout may learn patterns that yield better layouts through means of neural networks or genetic algorithms, but it will never think of copying best patterns created by another, similar robot that operates within its field of view - won't ever seek new sources of information.
    – SF.
    Jun 3, 2014 at 9:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .