It's tough, but try not to do it.
Labelling theory, put crudely, says that problems are only problems because someone — usually of course those in power, like teachers — define or label them as such (which does not of course deny that this may be a perfectly reasonable thing to do). So using a mobile phone in class is not inherently "bad", until someone labels it as such. And of course "answering back" to a teacher is only an offence—and a very subjective one—because teachers have the power to make it so.
The label of "deviant" (or perhaps something more benign) is then attached to people, so they may accept and internalise it, believing it of themselves. It is in this "secondary" sense which is most commonly used in educational contexts.
Becker and Lemert developed the theory, Hargreaves et al showed how it could apply in school classrooms, and Rosenthal and Jacobson suggested (in what is now regarded as a provocative but flawed study) that it could create a self-fulfilling prophecy in school such that children defined as "bright" would live up to expectations.
The stock-market is probably the primary field of self-fulfilling prophecy, in which if it is predicted that a stock will fall, there is a crisis of confidence and shareholders sell it, so that it does indeed fall in value (and don't we know it!) It can work the other way, too, as in the dot.com bubble.
In education, despite the Rosenthal and Jacobson study, labelling-based self-fulfilling prophecies usually operate to the detriment of students. Whole categories of students, based on gender, ethnicity or background, may be written off as incapable of achieving, setting up a frame of reference in which their failings are noticed and their achievements discounted. Individual students may also be labelled by being told they will never amount to anything, or that they are "no good at maths". Internalised, these labels are carried into new situations, including further and higher education, and the student is set up to fail.
There are obvious limits to the extent to which "upward" labelling can work, but few limits in the downward direction, and it is easy to fall into the trap. Unfortunately, if a student has been labelled in the past, she or he is likely to find that messages which reinforce that judgement come over much more clearly than any which contradict it. Casual remarks may be just as potent as formal assessments — or even more so, given that students are well aware of the formal constraints placed on recorded judgements.
- Be fair and respectful in accepting contributions in class: find something good in them even if they are not promising (without misleading students into believing they have the right answer when they do not, of course).
- Ensure your judgements do not go beyond the evidence: this piece of work was not up to the requirements, but that does not mean the student is hopeless. Eventually you may have to make that judgement, and this is not about making excuses for students, just about not generalising beyond the evidence.
- I hate to say it, but be sceptical about your colleagues' judgements, unless you have first-hand evidence to support them. Apply the same rigour to evidence in teaching as you would in your research.
BECKER H (1963) Outsiders;
studies in the sociology of deviance Glencoe,
Ill: Free Press
HARGREAVES D, HESTAR S and MELLOR F (1975) Deviance in Classrooms London; Routledge and Kegan Paul
LEMERT E (1972) Human Deviance, Social Problems and Social Control (2nd edn.) Englewood Cliffs N J; Prentice-Hall
ROSENTHAL R and JACOBSON L (1968) Pygmalion in the Classroom New York; Holt, Rinehart and Winston
OK: the references are ancient, but there are such things as "golden oldies".